JP2006185219A - Predicate logic formula creation device, summary journalizing rule creation device, summary journalizing device, predicate logic formula creation program, summary journalizing rule creation program and summary journalizing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a predicate logic formula creation device and a predicate logic formula creation program that create predicate logic formulas as rules of classifying data on each event by predetermined attributes, for events having background knowledge such as prerequisites for the events; a summary journalizing rule creation device and a summary journalizing rule creation program that create predicate logic formulas for events of accounting; and a summary journalizing device and a summary journalizing program that perform summary journalizing using predicate logic formulas. <P>SOLUTION: A summary journalizing rule creation process 30 applies common rules 32 of accounting transactions and company characteristic rules 33 specific to a company to past journalizing results 31 to create, by rough set theory, predicate logic formulas 34 whose decision attributes are a debit account title and a credit account title. A journalizing process 36 journalizes transaction data 35 according to the predicate logic formulas 34 to decide debit account titles and credit account titles and thus create journalizing results 37. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a predicate logical expression creating rule and predicate logical expression creating program that is a rule for classifying data for each event based on a predetermined attribute with respect to an event having a plurality of attributes. Predicate formula creation device, accounting journal rule creation device, accounting journal creation device, predicate formula creation program, accounting journal rule creation showing rules that determine what value other attributes will take when a value takes a predetermined value The present invention relates to an accounting journal program.

Conventionally, a method for creating a rule for classifying an event from a database storing real-world event data in a format that can be used for inference by a computer has been studied. For example, there is induction logic programming as a method for creating a predicate logic form rule as a hypothesis by using background knowledge such as preconditions about the event and examples (see Patent Document 1). Further, as a theory for inferring a logical relationship between sets with ambiguity, there is a rough set theory (see Patent Document 2).
Koichi Furukawa, Ken Ueno, Tomonobu Ozaki, “Inductive Logic Programming”, Kyoritsu Shuppan, May 25, 2001 Pawlak, Z .; "Rough Sets: Theoretical Aspects of Reasoning About Data", Kluwer Academic Publishers, 1991.

  However, there is a problem that an appropriate predicate logical expression cannot be generated from raw data constituting a fact only by induction logic programming as shown in Patent Document 1, and rough set logic as shown in Patent Document 2 However, there was a problem that an effective predicate logic form could not be created for cases with background knowledge.

  The present invention has been made to solve the above-mentioned problems. For events having background knowledge such as preconditions for the events, each event based on predetermined attributes using rough set theory. A predicate logical expression creation device for creating a predicate logical expression that is a rule for classifying data on the data, a predicate logical expression creation program, an accounting journal rule creation device for creating a predicate logical expression whose event is accounting processing, and It is an object of the present invention to provide an accounting journal rule creation program, an accounting journal device that performs accounting journal processing using a predicate logical expression, and an accounting journal program.

  In order to solve the above-mentioned problem, in the predicate logical expression creation device according to the first aspect of the present invention, for an event having a plurality of attributes, a predicate logic that is a rule for classifying data for each event based on a predetermined attribute In the predicate logical expression creation device for creating an expression, an event that is individual data for each event having at least a determination attribute that is an attribute serving as a classification reference and a condition attribute that is an attribute other than the determination attribute in a data item Event data input means for inputting data, event data storage means for storing a plurality of the event data input by the event data input means, and predetermined event data stored in the event data storage means For the condition attribute of the A cluster creating means for classifying the first cluster that is a combination and a second cluster that is a set of at least a part of the event data not included in the cluster, and the cluster created by the cluster creating means When the value of the predetermined condition attribute of the event data included in the first cluster is a numerical value, the value of the condition attribute of all the event data stored in the event data storage means A set having a value that is greater than or equal to the minimum value and less than or equal to the maximum value as a first condition attribute value set, and when the value of the predetermined condition attribute is a character string, a set having the value of each predetermined condition attribute as an element A first condition attribute value set creating means having the first condition attribute value set as a first condition attribute value set, and the predetermined condition of the event data included in the second cluster created by the cluster creating means If the value of the attribute is a numerical value, a set including elements whose values are the minimum value and the maximum value of the condition attribute values of all the event data stored in the event data storage means A second condition attribute value set, and when the value of the predetermined condition attribute is a character string, a second condition attribute value set is a set having the values of the predetermined condition attributes as elements The creation means compares the elements of the first condition attribute value set with the elements of the second condition attribute value set, and determines the relationship between the elements of the first condition attribute value set and the second condition attribute value set. Based on the determination result of the relationship determination means and the relationship determination means, a predetermined subset of the first condition attribute value set is set as a lower approximation set, or a predicate logical expression is created without creating a lower approximation set. Lower approximate set determining means not to create, and lower approximate set determining means A lower approximation set storage means for storing the lower approximation set determined by the step (a), and the lower approximation set stored in the lower approximation set storage means for the predetermined data of the event data constituting the lower approximation set. Based on the type of condition attribute, the value of the predetermined condition attribute, and the value of the determination attribute, a predicate logical expression indicating which value the determination attribute takes when the predetermined condition attribute takes a predetermined value A predicate logical expression creating means to be created and a predicate logical expression storage means for storing the previous description word logical expression created by the predicate logical expression creating means are provided.

  Further, in the predicate logical expression creation device of the invention according to claim 2, in addition to the configuration of the invention of claim 1, the relation judging means includes the element of the first condition attribute value set and the second condition attribute. The elements of the value set are compared, and the relationship between the elements of the first condition attribute value set and the second condition attribute value set is such that the first condition attribute value set and the second condition attribute value set overlap. The first relationship that does not have, the first condition attribute value set and the second condition attribute value set overlap each other and do not include each other, the second condition attribute value set is the first condition attribute value set Or the first condition attribute value set is a fourth relation included in the second condition attribute value set, and the approximate set determining means If the first condition is determined by the means, the first condition A first approximation set determining unit having a sex value set as a lower approximation set, and an element included in the overlap from the first condition attribute value set when the relationship determination unit determines that the second relationship is established In the case where it is determined that the third relation is determined by the second approximate set determining means that uses the excluded set as a lower approximate set and the relation determining means, the value of the predetermined condition attribute is a character string. When the set obtained by excluding the second condition attribute value set from the first condition attribute value set is a lower approximation set, and the value of the predetermined condition attribute is a numerical value, the maximum of the first condition attribute value set Of the two values adjacent to each other in the magnitude relation, the smaller value and the larger value of the value and the minimum value and the value of the second condition attribute value set included in the first condition attribute value set Before being included in the range indicated by the value Calculating the number of elements of the first condition attribute value set, and a third approximate set determining unit that sets a set of numbers included in a range specified by two adjacent values having the largest number as a lower approximate set; When it is determined by the relationship determining means that the relationship is the fourth relationship, and the value of the predetermined condition attribute is a numerical value, the maximum value and the minimum value of the first condition attribute value set are overlapped with the overlap. Among the values of the second condition attribute value set included, the first condition attribute value set included in a range indicated by a smaller value and a larger value of two adjacent values of all magnitude relationships A fourth approximation set determining unit that calculates a set of numbers included in a range specified by two adjacent values having the largest number as the lower approximation set; and When it is determined that the relationship is 4 In addition, when the value of the predetermined condition attribute is a character string, a non-creation unit that does not create a predicate logical expression is provided.

  Further, in the predicate logical expression creation device of the invention according to claim 3, in addition to the configuration of the invention of claim 1 or 2, there are a plurality of the condition attributes before being stored in the event data storage means. All condition attribute combination creating means for creating all combinations of the condition attributes of the event data, all condition attribute combination storage means for storing the combinations created by the all condition attribute combination creating means, and all condition attribute combinations For all combinations stored in the storage means, the cluster creation means, the first condition attribute value set creation means, the second condition attribute value set creation means, the relation determination means, and the lower approximation set determination, respectively. And an all-condition attribute combination processing means for executing the predescription word logical expression creating means.

  In addition to the configuration of the invention according to any one of claims 1 to 3, the predicate logical expression creation device according to the invention of claim 4 is an upper approximation set in which the first condition attribute value set is an upper approximation set. Creating means, an upper approximate event data number that is the number of event data in which the value of the predetermined condition attribute is included in the upper approximate set stored in the approximate set storage means, and the lower approximate set Calculate the lower approximate event data number that is the number of the event data including the value of the predetermined condition attribute, and calculate the approximate accuracy that is a value obtained by dividing the lower approximate event data number by the upper approximate event data number For each combination, the number of lower approximate event data of the combination, the approximate accuracy calculated by the approximate accuracy calculation means, and the event data stored in the event data storage means Based on the number of The effective performance evaluation index calculation means for calculating the effective performance evaluation index and the combination in which the value of the effective performance evaluation index calculated by the effective performance evaluation calculation means is equal to or less than a predetermined value A predicate logical expression creation control means for preventing the predescription word logical expression from being created in the expression creating means is provided.

  Further, in the predicate logical expression creation device of the invention according to claim 5, in addition to the configuration of the invention according to any one of claims 1 to 4, in the event data stored in the event data storage means If the value of the predetermined condition attribute is not stored, the cluster creation means, the first condition attribute value set creation means, the second condition attribute value set creation means, the relationship determination means, the lower approximation The set determining means and the predescription word formula creating means are not operated.

  Further, in the accounting journal rule creating device of the invention according to claim 6, the predicate formula creating device according to any one of claims 1 to 5, wherein the event is a transaction in accounting processing, and the event The data is transaction data relating to individual transactions that occur in accounting processing, the types of decision attributes are debits and credits, the elements thereof are account items, and the condition attributes are various related items relating to accounting transactions, The element is content related to a related item, and when a predetermined condition attribute takes a predetermined value, a predicate logical formula for determining which account item the debit and credit that are the determination attributes will be created It is the structure characterized by this.

  In the accounting journalizing apparatus of the invention according to claim 7, the transaction data input means for inputting the transaction data relating to each transaction generated in the accounting process, and the transaction data input by the transaction data input means are claimed in claim 6. According to the predescription word formula created by the predescription word formula creation means of the accounting journal rule creation device described in the above, the journal entry means for determining the debit and credit account items as the decision attributes, and the journal means Journal entry result storage means for storing the determined debit and credit account items that are the determined attributes in association with the transaction data is provided.

  In the predicate logical expression creation program of the invention according to claim 8, an operation for creating a predicate logical expression that is a rule for classifying data for each event based on a predetermined attribute for an event having a plurality of attributes Is a predicate formula creation program for causing a computer to execute at least a decision attribute that is an attribute that serves as a classification criterion and a conditional attribute that is an attribute other than the decision attribute in the data item. Event data input step for inputting event data that is data of the event data, an event data storage step for storing a plurality of the event data input by the event data input step in event data storage means, and the event data storage means For the predetermined condition attribute of the event data stored in the event data, A first cluster that is a set of at least some data of the event data that are the same, and a second cluster that is a set of at least some data of the event data not included in the cluster, If the value of the predetermined condition attribute of the event data included in the first cluster created by the cluster creation step is a numerical value, the cluster data is stored in the event data storage means. A set of elements having values from the minimum value to the maximum value of the condition attribute values of all the event data stored as a first condition attribute value set, and the value of the predetermined condition attribute is a character string In some cases, a first condition attribute value set creation step in which a set having the values of the predetermined condition attributes as elements is a first condition attribute value set; and the cluster creation If the value of the predetermined condition attribute of the event data included in the second cluster created by the step is a numerical value, all of the event data stored in the event data storage means A set of elements having values from the minimum value to the maximum value of the condition attribute value as a second condition attribute value set, and when the value of the predetermined condition attribute is a character string, A second condition attribute value set creation step in which a set having values as elements is a second condition attribute value set, and comparing the elements of the first condition attribute value set with the elements of the second condition attribute value set; A relationship determination step of determining a relationship between elements of one condition attribute value set and the second condition attribute value set; and a predetermined subset of the first condition attribute value set based on a determination result in the relationship determination step. Lower approximation set or lower A lower approximation set determining step that does not create a predicate logical expression without creating an approximate set, and a lower approximate set storage step that stores the lower approximate set determined by the lower approximate set determination step in a lower approximate set storage means And for the lower approximation set stored in the lower approximation set storage means, the type of the predetermined condition attribute of the event data constituting the lower approximation set, the value of the predetermined condition attribute, and the determination attribute A predicate logical expression creating step for creating a predicate logical expression indicating which value the decision attribute takes when the predetermined conditional attribute takes a predetermined value based on the value of the predicate logical expression, A predicate logical expression storage step for storing the created predescription logical expression in the predicate logical expression storage means is executed by a computer.

  Further, in the predicate logical expression creation program of the invention according to claim 9, in addition to the configuration of the invention according to claim 8, the relationship determination step includes: The elements of the two condition attribute value sets are compared, and the relationship between the elements of the first condition attribute value set and the second condition attribute value set is the following: A first relationship that does not have an overlap, a second relationship in which the first condition attribute value set and the second condition attribute value set do not overlap each other, and the second condition attribute value set is the first condition attribute Determining whether the third relation included in the value set or the first relation attribute value set is the fourth relation included in the second condition attribute value set, and the approximate set determination step includes: In the relationship determination step, the first relationship Is included in the overlap when it is determined that the second relation is determined in the first approximate set determining step in which the first condition attribute value set is a lower approximate set and the relation determining step. A second approximate set determining step in which a set obtained by removing the element to be removed from the first condition attribute value set is a lower approximate set, and the predetermined condition when the relationship determining step determines that the third relationship is satisfied When the attribute value is a character string, a set obtained by removing the second condition attribute value set from the first condition attribute value set is set as a lower approximation set, and the value of the predetermined condition attribute is a numerical value. Are the two adjacent values of all the magnitude relationships in the maximum value and the minimum value of the first condition attribute value set and the value of the second condition attribute value set included in the first condition attribute value set. Its value small The number of elements of the first condition attribute value set included in the range indicated by the larger value and the larger value is calculated, and the number included in the range specified by the two adjacent values having the largest number If the value of the predetermined condition attribute is a numerical value when it is determined in the third approximate set determining step that is a lower approximate set and the relationship determining step is the fourth relationship, Of the maximum value and the minimum value of the first condition attribute value set and the value of the second condition attribute value set included in the overlap, the smaller value and the larger value of two adjacent values of all magnitude relations The number of elements of the first condition attribute value set included in the range indicated by the value of is calculated, and the set of numbers included in the range specified by the two adjacent values having the largest number is calculated as a lower approximation set The fourth approximate set decision step And a non-creation step that does not create a predicate logical expression when the value of the predetermined condition attribute is a character string when it is determined that the relationship is the fourth relationship in the relationship determination step; Is executed by a computer.

  Further, in the predicate logical expression creation program of the invention according to claim 10, in addition to the configuration of the invention according to claim 8 or 9, there are a plurality of the condition attributes before the condition data is stored in the event data storage means. All condition attribute combination creating step for creating all combinations of the condition attributes of the event data, and all condition attribute combination storing step for storing the combination created by the all condition attribute combination creating step in the all condition attribute combination storage means And the cluster creation step, the first condition attribute value set creation step, the second condition attribute value set creation step, and the relationship determination for all combinations stored in the all condition attribute combination storage means, respectively. A step, a lower approximation set determination step and a predescription word formula creation step. It has a structure, characterized in that to execute the computer.

  In addition to the configuration of the invention according to any one of claims 8 to 10, the predicate formula creating program of the invention according to claim 11 is an upper approximation set in which the first condition attribute value set is an upper approximation set. A creation step, an upper approximate event data number that is the number of event data in which the value of the predetermined condition attribute is included in the upper approximate set stored in the approximate set storage means, and the lower approximate set Calculate the lower approximate event data number that is the number of the event data including the value of the predetermined condition attribute, and calculate the approximate accuracy that is the value obtained by dividing the lower approximate event data number by the upper approximate event data number For each combination, the number of lower approximation event data of the combination, the approximation accuracy calculated by the approximation accuracy calculation step, and the event data storage means before being stored in the event data storage means An effective performance evaluation index calculating step for calculating an effective performance evaluation index based on the number of event data, and the combination in which the value of the effective performance evaluation index calculated by the effective performance evaluation calculating step is equal to or less than a predetermined value Is configured to cause the computer to execute a predicate logical expression creation control step for preventing the pre-description logical expression from being created in the previous description word logical expression creation step.

  Further, in the predicate logical expression creation program of the invention according to claim 12, in addition to the configuration of the invention of any one of claims 8 to 11, in the event data stored in the event data storage means If the value of the predetermined condition attribute is not stored, the cluster creation step, the first condition attribute value set creation step, the second condition attribute value set creation step, the relationship determination step, the lower approximation The set determination step and the pre-description word formula creation step are prevented from being executed by a computer.

  Further, in the accounting rule creation program of the invention according to claim 13, in the predicate formula creation program according to any one of claims 8 to 12, the event is a transaction in accounting processing, and the event data is Transaction data relating to individual transactions that occur in accounting processing, the types of decision attributes are debits and credits, elements thereof are account items, and the condition attributes are various related items relating to accounting transactions, Is the contents related items, and when a predetermined condition attribute takes a predetermined value, the computer creates a predicate logical expression that determines which account item the debit and credit that are the determination attributes will be It is the structure characterized by making it carry out.

  In the accounting journal program according to the fourteenth aspect of the present invention, the transaction data input step for inputting transaction data relating to each transaction generated in the accounting process, and the transaction data input by the transaction data input step are defined in claim 13. According to the predescription word formula created by the predescription word formula creation step of the accounting journal rule creation program described in the above, the journal entry step for determining the debit and credit account items as the decision attributes, and the journal step The journalizing result storing step for storing the determined debit and credit account items, which are the determined attributes, in association with the transaction data is executed by a computer.

  In the predicate logical expression creation device of the invention according to claim 1, the event data input means has at least a decision attribute that is an attribute that serves as a classification criterion and a condition attribute that is an attribute other than the decision attribute in each data item. Event data which is individual data about the event data is input, and the event data storage means can store a plurality of event data input by the event data input means. Then, the cluster creating means determines the event data for the predetermined condition attribute of the event data stored in the event data storage means, and a set of at least a part of the event data having the same attribute value The first condition attribute value set creation means is created by the cluster creation means, and is classified into a second cluster that is a set of at least a part of event data not included in the cluster. When the value of the predetermined condition attribute of the event data included in the first cluster is a numerical value, a value not less than the minimum value and not more than the maximum value of the condition attribute values of all event data stored in the event data storage means Is a first condition attribute value set, and when the value of the predetermined condition attribute is a character string, the set having the value of each predetermined condition attribute as the element The condition attribute value set is set, and the second condition attribute value set creating means is the event data storage means when the value of the predetermined condition attribute of the event data included in the second cluster created by the cluster creating means is a numerical value. When the second condition attribute value set is a set of elements whose values are not less than the minimum value and not more than the maximum value of the condition attribute values of all event data stored in, and the value of the predetermined condition attribute is a character string Can set a set of elements having the values of each predetermined condition attribute as a second condition attribute value set. Then, the relationship determination means compares the elements of the first condition attribute value set and the elements of the second condition attribute value set, and determines the relationship between the elements of the first condition attribute value set and the second condition attribute value set. The lower approximate set determining means sets a predetermined subset of the first condition attribute value set as the lower approximate set based on the determination result of the relationship determining means, or predicate logical expression without creating the lower approximate set The lower approximation set storage means stores the lower approximation set determined by the lower approximation set determination means, and the predicate logical expression creation means stores the lower approximation set stored in the lower approximation set storage means. Which value the decision attribute takes when the predetermined condition attribute takes a predetermined value based on the type of the predetermined condition attribute of the event data constituting the lower approximation set, the value of the predetermined condition attribute, and the value of the decision attribute Create a predicate formula that indicates whether Means can store the predicate logic formulas created by predicate logic formula creation means. Therefore, since a rule (predicate logical expression) for classifying data related to a certain event with respect to a determination attribute can be created, the determination attribute of data related to a certain event can be easily determined by using the created predicate logical expression. It becomes like.

  Further, in the predicate logical expression creation device according to the second aspect of the invention, in addition to the effect of the first aspect of the invention, the relationship judging means includes the elements of the first condition attribute value set and the second condition attribute value set. The elements are compared, and the element relationship between the first condition attribute value set and the second condition attribute value set is the first relation in which the first condition attribute value set and the second condition attribute value set do not overlap, The first condition attribute value set and the second condition attribute value set are overlapped and not included in each other, the second condition attribute value set is included in the first condition attribute value set, or the first relation It is determined whether the condition attribute value set is one of the fourth relations included in the second condition attribute value set, and the first approximate set determining means of the approximate set determining means is the first relation in the relation determining means. If it is determined, the first condition attribute value set is set as the lower approximation set, and the second approximation set determiner When the relationship determining means determines that the second relationship is established, the set obtained by removing the elements included in the overlap from the first condition attribute value set is set as the lower approximate set, and the third approximate set determining means is the relationship determining When it is determined that the third relationship is established by the means, and the value of the predetermined condition attribute is a character string, a set obtained by removing the second condition attribute value set from the first condition attribute value set is a lower approximation set. When the value of the predetermined condition attribute is a numerical value, the maximum value and the minimum value of the first condition attribute value set, and the value of the second condition attribute value set included in the first condition attribute value set , The number of elements of the first condition attribute value set included in the range indicated by the smaller value and the larger value of two adjacent values of all magnitude relations is calculated, and the number is the largest. Included in the range specified by two adjacent values And the fourth approximate set determining means determines that the first condition attribute is a numerical value when the value of the predetermined condition attribute is a numerical value when the relation determining means determines that the relation is the fourth relation. In the maximum value and the minimum value of the value set and the value of the second condition attribute value set included in the overlap, it is indicated by the smaller value and the larger value of two adjacent values of all magnitude relations. The number of elements of the first condition attribute value set included in the range is calculated, the set of numbers included in the range specified by two adjacent values having the largest number is set as a lower approximation set, When the relationship determining means determines that the relationship is the fourth relationship, if the value of the predetermined condition attribute is a character string, the predicate logical expression may not be created. Therefore, a set of condition attribute values of elements of the set having the same decision attribute value (first condition attribute value set) and a set of condition attribute values of elements of the other set of decision attribute values (second condition attribute value set) ), The lower approximation set can be determined in correspondence with whether the attribute value is a character string or a numerical value.

  Further, in the predicate logical expression creation device of the invention according to claim 3, in addition to the effect of the invention of claim 1 or 2, since there are a plurality of condition attributes, all condition attribute combination creation means All combinations of the condition attributes of the event data stored in the storage means are created, and the all condition attribute combination storage means can store the combinations created by the all condition attribute combination creation means. Then, the all condition attribute combination processing means has a cluster creation means, a first condition attribute value set creation means, and a second condition attribute value set creation means for all combinations stored in the all condition attribute combination storage means. , Relationship determination means, lower approximation set determination means, and predicate logical expression creation means. Therefore, a predicate logical expression can be created using various condition attributes related to events as conditions and combinations thereof as conditions.

  In addition, in the predicate logical expression creation device of the invention according to claim 4, in addition to the effect of the invention according to any one of claims 1 to 3, the upper approximation set creation means further approximates the first condition attribute value set. The approximate accuracy calculation means is a set of upper approximate event data, which is the number of event data in which the value of the predetermined condition attribute is included in the upper approximate set stored in the approximate set storage means, and the lower approximate set. The number of event data including the value of the condition attribute is calculated as the lower approximate event data number, the approximate accuracy is calculated by dividing the lower approximate event data number by the upper approximate event data number, and the effective performance evaluation index The calculation means, for each combination, the effective performance evaluation index based on the number of lower approximate event data of the combination, the approximation accuracy calculated by the approximation accuracy calculation means, and the number of event data stored in the event data storage means Calculate the predicate logic The creation control means can prevent the predicate logical expression creation means from creating a predicate logical expression for a combination in which the value of the effective performance evaluation index calculated by the effective performance evaluation calculation means is equal to or less than a predetermined value. . Therefore, since a predicate logical expression with low effectiveness is not created, an invalid classification is not performed when event data is classified using the created predicate logical expression.

  In addition, in the predicate logical expression creation device of the invention according to claim 5, in addition to the effect of the invention according to any one of claims 1 to 4, in the event data stored in the event data storage means, a predetermined condition If the attribute value is not stored, the cluster creation means, the first condition attribute value set creation means, the second condition attribute value set creation means, the relationship judgment means, the lower approximation set determination means, and the predicate logical expression creation means It can be prevented from operating. Therefore, since no unnecessary processing is performed on data having no condition attribute value, the processing load can be reduced and the processing time can be shortened.

  In addition, in the accounting journal rule creation device of the invention according to claim 6, in addition to the effect of the invention according to any one of claims 1 to 5, an event is a transaction in accounting processing, and event data is generated in accounting processing Transaction data related to individual transactions, debit and credit as the type of decision attributes, elements as account items, condition attributes as various related items related to accounting transactions, elements as contents related items, predetermined condition attributes When taking a predetermined value, it is possible to create a predicate logical expression that determines which account item the debit and credit that are the determination attributes are. Therefore, based on the past journal results (event data), it is possible to create a predicate formula for the journal processing of determining the debit account and credit account in the accounting process. Rules (predicate logical expressions) can be created.

  In the accounting journalizing apparatus of the invention according to claim 7, the transaction data input means inputs the transaction data relating to each transaction generated in the accounting process, and the journalizing means receives the transaction data input by the transaction data input means. The debit and credit account items that are decision attributes are determined according to the predicate logical expression created by the predicate logical expression creating means of the accounting journal rule creating device according to claim 6, and the journal result storage means is determined by the journalizing means. The debit and credit account items, which are the determined attributes, can be stored in association with the transaction data. Therefore, the transaction data can be easily journaled by the accounting journal rule (predicate logical formula) corresponding to the company.

  Further, in the predicate logical expression creation program of the invention according to claim 8, for each event, the computer has at least a determination attribute that is an attribute serving as a classification reference and a condition attribute that is an attribute other than the determination attribute in a data item. An event data input step for inputting event data that is individual data of the event data, an event data storage step for storing a plurality of event data input by the event data input step in the event data storage means, and an event data storage means The event data that is not included in the cluster is the first cluster that is a set of at least a part of the event data having the same value as the determination attribute value for the predetermined condition attribute of the event data. Cluster creation step for classifying at least part of the data into a second cluster that is a set of data If the value of the predetermined condition attribute of the event data included in the first cluster created by the cluster creation step is a numerical value, the value of the condition attribute of all the event data stored in the event data storage means A set of elements having a value between the minimum value and the maximum value of the first is the first condition attribute value set. When the value of the predetermined condition attribute is a character string, the set having the value of each predetermined condition attribute as the element If the value of the predetermined condition attribute of the event data included in the second cluster created by the first condition attribute value set creation step and the cluster creation step as the first condition attribute value set is a numerical value, the event data A set of elements having values from the minimum value to the maximum value of the condition attribute values of all event data stored in the storage means as the second condition attribute value set, and the value of the predetermined condition attribute is a character string A second condition attribute value set creating step in which a set having each predetermined condition attribute value as an element is a second condition attribute value set, an element of the first condition attribute value set, and a second condition attribute value set And a relationship determination step for determining the relationship between the elements of the first condition attribute value set and the second condition attribute value set, and based on the determination result at the relationship determination step, The lower approximation set determined by the lower approximation set determining step and the lower approximation set determining step in which the predetermined subset is a lower approximation set, or the predicate logical expression is not created without creating the lower approximation set. The lower approximation set storage step stored in the lower approximation set storage means, and the lower approximation set stored in the lower approximation set storage means, the types of predetermined condition attributes of event data constituting the lower approximation set, the predetermined condition attributes Values and determinants The predicate formula creation step and the predicate formula creation step create a predicate formula that indicates what value the decision attribute takes when a given condition attribute takes a given value based on the sex value The predicate logical expression storage step of storing the predicate logical expression in the predicate logical expression storage means can be executed. Therefore, since a rule (predicate logical expression) for classifying data related to a certain event with respect to a determination attribute can be created, the determination attribute of data related to a certain event can be easily determined by using the created predicate logical expression. It becomes like.

  Further, in the predicate logical expression creation program of the invention according to claim 9, in addition to the effect of the invention of claim 8, the relationship determination step causes the computer to send the elements of the first condition attribute value set and the second condition attribute value. The elements of the set are compared, and the relationship between the elements of the first condition attribute value set and the second condition attribute value set is the first relation in which the first condition attribute value set and the second condition attribute value set do not overlap. , A second relationship in which the first condition attribute value set and the second condition attribute value set overlap and are not included in each other, a third relationship in which the second condition attribute value set is included in the first condition attribute value set, or It can be determined whether the first condition attribute value set is one of the fourth relationships included in the second condition attribute value set. In the approximate set determining step, when it is determined that the computer has the first relationship in the relationship determining step, in the first approximate set determining step in which the first condition attribute value set is the lower approximate set, and in the relationship determining step When it is determined that the relationship is the second relationship, a second approximate set determining step in which a set obtained by removing the elements included in the overlap from the first condition attribute value set is a lower approximate set; If it is determined that the value of the predetermined condition attribute is a character string, a set obtained by removing the second condition attribute value set from the first condition attribute value set is set as a lower approximation set, and the predetermined condition attribute is set. When the value of is a numerical value, all the magnitude relations between the maximum and minimum values of the first condition attribute value set and the values of the second condition attribute value set included in the first condition attribute value set Next to Calculates the number of elements of the first condition attribute value set included in the range indicated by the smaller value and the larger value for the two matching values, and is specified by the two adjacent values with the largest number When the value of the predetermined condition attribute is a numerical value when it is determined in the third approximate set determining step that sets the number of sets included in the range to be the lower approximate set and the relationship determining step is the fourth relationship Is the smaller value and the larger value of the two adjacent values of all magnitude relationships in the maximum and minimum values of the first condition attribute value set and the values of the second condition attribute value set included in the overlap. The number of elements of the first condition attribute value set included in the range indicated by the value of, and the set of numbers included in the range specified by the two adjacent values having the largest number is the lower approximation set The fourth approximate set decision step And a non-creation step in which a predicate logical expression is not created when the value of the predetermined condition attribute is a character string when it is determined that the relationship is the fourth relationship in the relationship determination step. Can do. Therefore, a set of condition attribute values of elements in the set having the same decision attribute value (first condition attribute value set) and a set of condition attribute values of elements of the other set of decision attribute values (second condition attribute value set) ), The lower approximation set can be determined in correspondence with whether the attribute value is a character string or a numerical value.

  Further, in the predicate logical expression creation program of the invention according to claim 10, in addition to the effect of the invention of claim 8 or 9, all the condition attributes of event data stored in the event data storage means are stored in the computer. All condition attribute combination creating step for creating a combination of all conditions, all condition attribute combination storing step for storing the combination created by the all condition attribute combination creating step in all condition attribute combination storing means, and storing in all condition attribute combination storing means A cluster creation step, a first condition attribute value set creation step, a second condition attribute value set creation step, a relationship determination step, a lower approximation set determination step, and a predicate logic formula creation step are performed for all combinations that are performed. All condition attribute combination processing steps to be executed can be executed. Therefore, a predicate logical expression can be created using various condition attributes related to events as conditions and combinations thereof as conditions.

  In addition to the effects of the invention according to any one of claims 8 to 10, the predicate logical expression creation program according to the invention according to claim 11 is used to make the computer set the first condition attribute value set as an upper approximation set. Approximate set creation step, upper approximate event data number that is the number of event data whose value of the predetermined condition attribute is included in the upper approximate set stored in the approximate set storage means, and predetermined condition attribute in the lower approximate set For each combination, an approximate accuracy calculation step for calculating the number of lower approximate event data, which is the number of event data including the value of, and calculating the approximate accuracy that is the value obtained by dividing the number of lower approximate event data by the number of upper approximate event data, In addition, an effective performance evaluation index is calculated based on the number of lower approximate event data in the combination, the approximate accuracy calculated in the approximate accuracy calculation step, and the number of event data stored in the event data storage means. Do not create a predicate logical expression in the predicate logical expression creation step for a combination in which the value of the effective performance evaluation index calculation step and the effective performance evaluation index calculated by the effective performance evaluation calculation step is equal to or less than a predetermined value. And a predicate formula creation control step. Therefore, since a predicate logical expression with low effectiveness is not created, an invalid classification is not performed when event data is classified using the created predicate logical expression.

  Further, in the predicate logical expression creation program of the invention according to claim 12, in addition to the effect of the invention of any one of claims 8 to 11, in the event data stored in the event data storage means, a predetermined condition If the attribute value is not stored, a cluster creation step, a first condition attribute value set creation step, a second condition attribute value set creation step, a relationship determination step, a lower approximation set determination step, and a predicate logical expression creation step are performed. You can prevent the computer from implementing it. Therefore, since no unnecessary processing is performed on data having no condition attribute value, the processing load can be reduced and the processing time can be shortened.

  In the accounting rule creation program of the invention according to claim 13, an event in the predicate formula creation program according to any one of claims 8 to 12 is regarded as a transaction in accounting processing, and event data is generated in accounting processing. The transaction data related to the transaction is debit and credit, the element is the account item, the condition attribute is various related items related to the accounting transaction, and the element is the content related to the specified condition attribute. When taking a predetermined value, it is possible to cause the computer to create a predicate logical expression that determines which account item the debit and credit as the determination attributes are. Therefore, based on the past journal results (event data), it is possible to create a predicate formula for the journal processing of determining the debit account and credit account in the accounting process. Rules (predicate logical expressions) can be created.

  In the accounting journal program of the invention according to claim 14, the transaction data input step for inputting transaction data relating to each transaction generated in the accounting process, and the transaction data input by the transaction data input step are described in claim 13. The journal step that determines the debit and credit account items that are the decision attributes according to the predicate formula created by the predicate formula creation step of the accounting journal rule creation program, and the debit that is the decision attribute determined by the journal step And a journal result storing step for storing the credit account item in association with the transaction data. Therefore, the transaction data can be easily journaled by the accounting journal rule (predicate logical formula) corresponding to the company.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. First, data classification using predicate logical expressions will be described with reference to FIG. FIG. 1 is a schematic diagram showing the concept of data classification using predicate logical expressions. Various data related to the natural world and society (event data) is generated. And when it is going to take some action about the event, classification of event data is needed. In order to classify the event data, what is the basis of the event data is important.

  The event data is a collection of values from various viewpoints (attributes) regarding the event. For example, if you focus on accounting transactions in the accounting system, such as the date the transaction occurred, the purpose of the transaction (sales, salary, etc.), the amount, the type of money such as cash, check or bill, balance sheet, etc. Accounts when creating accounting materials correspond to attributes. Further, if focusing on student information at school, the name, gender, address, school of origin, past class, current class, test score, affiliation activity, etc. correspond to attributes.

  When classifying event data, there is a case where the value of an attribute (referred to as a “decision attribute”) that serves as a reference for classification is not determined. For example, in accounting transactions, in order to create accounting materials, it is necessary to classify by account item, but the account item is determined even if the date, purpose, amount, money type, etc. are known from the actual slip. May not have. The work of classifying event data generated from a slip into account items is called a journal and is a basic accounting process. Also, in the case of school, considering the case of class classification, the class becomes a determination attribute, but the purpose is to determine the class, and the event data class is undecided.

  Therefore, if the classification rule for the decision attribute can be created using another attribute value (referred to as “condition attribute”) as a condition, the event data can be easily classified by the decision attribute. In the present invention, a predicate logical expression serving as a rule for this classification is created. In FIG. 1, the rule (predicate logical expression) 12 corresponds to this. It should be noted that in the present invention, it is not premised on case data (negative examples) that are difficult to collect in an actual problem-solving field, that is, event data that does not appear to occur in the natural world or society.

  As shown in FIG. 1, it is assumed that seven event data of event data 1, 2, 3, 4, 5, 6, and 7 are obtained in event 10. In the event data 1 to 7, it is desired to classify a certain attribute (decision attribute), but the value is unknown. Therefore, in order to classify the decision attributes, rules (predicate logical expressions) 12 are used. The rule 12 is derived from past event data 13 in the event 10, a condition (common rule) 14 recognized by the expert for the event 10, and a special rule (special rule) 15 in the event 10. The process for deriving this rule 12 is the main part of the present invention. The generation of the rule 12 will be described in detail later with reference to FIGS. 2 to 10, and then described with reference to specific examples with reference to FIGS. 11 to 33.

  For example, if there are four condition attributes of attributes a, b, c, and d and one decision attribute, the rule 12 determines that “if the value of attribute a> 1, and the value of attribute b ≦ 6, The value of the attribute is “A”. ", If the value of attribute c is" B ", the value of the decision attribute is" A ". , “10 <attribute a and 6 <attribute b, the value of the determination attribute is“ B ”. , “100 <attribute d, the value of the determination attribute is“ B ”. ", If the value of attribute c is" A "or" F ", the value of the decision attribute is" C ". , “If attribute a ≦ 1, and attribute d ≦ 100, the value of the determination attribute is“ C ”. Etc.

  In the example illustrated in FIG. 1, according to the above rules, event data 1 is classified as “attribute a = 3, attribute b = 4.5” and determined attribute “A”, and event data 2 is “attribute a = 12, attribute b = 7 ”is classified as a decision attribute“ B ”, event data 3 is classified as“ attribute d = 125 ”and decision attribute“ B ”, and event data 4 is classified as“ attribute c is “ "A" "is classified as a decision attribute" C ", event data 5 is classified as" attribute a = 1, attribute d = 96 "as a decision attribute" C ", and event data 6 is classified as" attribute c " “B” is classified as a decision attribute “A”, and the event data 7 is classified as “decision attribute“ A ”with“ attribute a = 3, attribute b = 4.8 ”. Therefore, as shown in FIG. 1, the decision attributes of the event data 1 to 7 are decided, the decision attribute “A” set 16, the decision attribute “B” set 17, and the decision attribute “C” set 18. It is classified into three.

In the present embodiment, it is assumed that the rule 12 is expressed by a predicate logical expression in the form of Equation 1. The right side of “: −” is a condition, the left side is the result, and if the right side is true, the relationship that the left side is also true is established. For example, if the value of attribute a is X1> 1, and the value of attribute b is X2 ≦ 6, the value of the determination attribute is “A”. Is expressed as the following formula (2). Also, if the value Y1 of the attribute c is “A” or “F”, the value of the decision attribute is “C”. Is expressed as the following equation (3). “GlaerThanInt (X1,1)” is a notation indicating that X1 is an integer type and greater than 1. Correspondingly, “lessThanInt (X1,1)” is X1 being an integer type and less than 1, “lessEqualInt (X1,1)” is X1 being an integer type and less than or equal to 1, and “greaterThanInt (X1,1)” ) ”X1 is an integer type and indicates 1 or more. When X1 is a real type, it is set to “Float” instead of “Int”. “Member (Y1 [A, F])” is a notation indicating that the attribute value (variable Y1) is “A” or “F”, and the attribute value is listed in [].

  Next, a method for creating the rule 12 will be described with reference to FIGS. 2 is a schematic diagram of a decision table 20 used in rough set theory, FIG. 3 is a schematic flowchart of rule creation, and FIG. 4 is a schematic diagram of a partial decision table 21 for the condition attribute a of the decision table 20. FIG. FIG. 5 is an image diagram of a cluster generated from the partial determination table 21. 6 and 7 are image diagrams of the lower approximation set when the condition attribute takes a character string, and FIGS. 8 to 10 are image diagrams of the lower approximation set when the condition attribute takes a numerical value.

  As shown in FIG. 2, the event data to be classified is expressed by arranging the condition attribute and the decision attribute separately. Based on this decision table 20, a rule 12 is created in the flow of the flowchart shown in FIG.

  First, as shown in FIG. 2, the attribute of event data is divided into a decision attribute and a condition attribute, and a decision table 20 is created (S1). Then, a combination for each condition attribute is created (S2). In the condition attribute as shown in FIG. 2, (attribute a), (attribute b), (attribute c),..., (Attribute a, attribute b), (attribute a, attribute c), (attribute b, attribute c),... (attribute a, attribute b, attribute c),. And the process of S4-S9 is performed about all the combinations (S3: NO, S4-S9). First, in the combination of condition attributes to be processed, event data is classified into a first cluster and a second cluster for each value of the decision attribute (S4).

  Specifically, a partial determination table 21 (see FIG. 4) is created from the determination table 20 and including values of only the determination attribute and the attribute of the combination. At this time, event data for which no value is set in the attribute of the combination is deleted from the partial determination table 21. FIG. 4 is a schematic diagram of the partial determination table 21 in which the determination attribute is only the attribute α and the combination of the condition attributes is created for (attribute a). First, each data is sorted by the value of the decision attribute. In the example shown in FIG. 5, the four pieces of data from the top have a decision attribute (attribute α) of “1”, and the remaining two pieces of data have a decision attribute (attribute α) of “2”. Therefore, each decision attribute value is divided into a first cluster (a set of data having the same decision attribute value) and a second cluster (a set of data having a decision attribute value other than the decision attribute value of the first cluster). The In the example shown in FIG. 5, since there are two types of determination attribute α, “1” and “2”, two first clusters are created. Assuming (value of condition attribute a, value of decision attribute α), the first first cluster is {(12,1), (54,1), (12,1), (13,1)} The first second cluster is {(21, 2), (12, 2)}. The second first cluster is {(21,2), (12,2)}, and the second second cluster is {(12,1), (54,1), (12,1), (13, 1)}.

  And the process of S6-S9 is performed about the combination of all the decision attributes produced by S4 (S5: NO, S6-S9). First, an upper approximation set is created (S6), a lower approximation set is created (S7), and an effectiveness evaluation value is calculated based on the upper approximation set and the lower approximation set (S8). A predicate logical expression is created (S9). Then, when the processing is completed for all combinations of determined attributes (S5: YES), the processing is performed for the next combination of conditional attributes (S4 to S9), and when the processing for all combinations is completed (S3: YES). Then, the predicate logical expression whose effectiveness evaluation value is equal to or smaller than the predetermined value is deleted (S10), and the creation of the rule (predicate logical expression) from the event data is completed.

Here, a method of determining the upper approximation set and the lower approximation set will be described with reference to FIGS. The upper approximate set and the lower approximate set are sets used in the rough set theory. Equation 4 shows the lower approximation set, and Equation 5 shows the upper approximation set. In Equations 4 and 5, “X” indicates event data to be identified, “U” indicates an entire set of event data, “x” indicates one event data, “θ” indicates an empty set, and R (x ) Indicates a set of event data whose determination attribute is determined by the condition attribute R. R _* (X) represents the lower approximate set, and R ^* (X) represents the upper approximate set.

  That is, the lower approximate set indicates that all elements of the R (x) set are included in the X set, and all elements included in the lower approximate set have values of the determination attribute. The upper approximation set indicates that there is a common part between the R (x) set and the X set (the common part is not an empty set), and at least one of the elements of the R (x) set is X It is said to be an element of a set. If the decision attribute can be completely classified by the condition attribute R, the upper approximate set and the lower approximate set are the same set.

  The upper approximate set is the same set as the first cluster. The lower approximate set includes a set of condition attribute values of event data belonging to the first cluster (first condition attribute value set) and a set of condition attribute values of event data belonging to the second cluster (second condition attribute value). It is determined by the relationship with the set. In particular, when determining the lower approximation set, the determination method differs depending on whether the value of the condition attribute value is a character string or a numerical value that has a magnitude relationship between the values.

  First, a case where the condition attribute value is a character string will be described. Here, the first cluster is P, and the second cluster is Q. When there is no overlap between the first cluster P and the second cluster Q (P∩Q = empty set) (corresponding to the “first relationship”), if the value belonging to the first cluster P is used as a conditional attribute, it is always The decision attribute is determined. Therefore, both the upper approximate set and the lower approximate set are set as the first cluster P.

  Further, as shown in FIG. 6, when the first cluster P and the second cluster Q overlap and are not included in the other (P∩Q ≠ empty set, and P is not a subset of Q, and , Q is not a subset of P) (corresponding to “second relation”), when the value belonging to the overlap is a condition attribute, the value of the decision attribute is not uniquely determined. That is, if the first cluster P is a set of data having the determination attribute “1”, if the condition attribute value is “A” belonging to the overlap, the determination attribute is “1”, Some of them are not “1”. Therefore, the first cluster P is an upper approximation set, and a set (P− (P （Q)) obtained by deleting the overlap (product) of the first cluster P and the second cluster Q from the first cluster is a lower approximation set. Is done.

  Further, as shown in FIG. 7, when the second cluster Q is included in the first cluster P (Q⊆P) (corresponding to the “third relation”), the first cluster P and the second cluster Q According to the case where there is an overlap, the first cluster P is an upper approximate set, and the set obtained by deleting the product of the first cluster P and the second cluster Q from the first cluster (P− (P∩Q)) is lower. It is an approximate set.

  When the first attribute P is included in the second cluster Q when the condition attribute is a character string (P⊆Q) (corresponding to “fourth relationship”), the determination is made based on the value of the condition attribute. An attribute cannot be uniquely determined. That is, if the first cluster P is a set of data having the determination attribute “1”, if the condition attribute value is “A”, the determination attribute is “1” and not “1”. It means that there are things. That is, the value of the determination attribute cannot be determined from the value of the condition attribute. Therefore, it can be said that this event data is data that is not suitable for creating a predicate logical expression, and the predicate logical expression is not created.

  Next, a case where the condition attribute value is a number will be described. When the condition attribute takes a numerical value, if the first cluster P and the second cluster Q do not overlap (P∩Q = empty set) (corresponding to the “first relationship”), the condition attribute is a character string. If the value belonging to the first cluster P is used as the condition attribute as in the case of taking the decision attribute, the determination attribute is always determined. Therefore, both the upper approximate set and the lower approximate set are set as the first cluster P.

  Further, as shown in FIG. 8, when the first cluster P and the second cluster Q overlap and are not included in the other (P∩Q ≠ empty set, P is not included in Q, and Q Is not included in P) (corresponding to “second relation”), the first cluster P is an upper approximation set, and the product of the first cluster P and the second cluster Q is deleted from the first cluster (P− (P∩Q)) is a lower approximation set. In the example shown in FIG. 8, the regions of the first cluster P and the second cluster Q overlap in the range of q1 ≦ x ≦ p4. Therefore, the lower approximation set is p1 ≦ x <q1. Here, in the range specified by two values, if the value is an element of the first cluster P, an equal sign is included (≧, ≦), and if the value is an element of the second cluster Q, the equal sign is included. Not included (>, <).

  Further, as shown in FIG. 9, when the second cluster Q is included in the first cluster P (corresponding to the “third relationship”), the maximum value and the minimum value of the first cluster P, and the first cluster P And the value of the second cluster Q included in the overlap between the second cluster Q and the first cluster P included in the range of the smaller value and the larger value of the two adjacent values of the magnitude relationship And the set of elements of the first cluster P included in the range between the two adjacent values having the largest number is set as the lower approximation set. The relationship between the maximum value (p6) and the minimum value (p1) of the first cluster P and the values (q1, q2, q3) of the second cluster Q included in the first cluster P is p1 <q1 <q2 <q3. <P6. Therefore, in the five ranges R1 (p1 ≦ x ≦ q1), R2 (q1 ≦ x ≦ q2), R3 (q2 ≦ x ≦ q3), R4 (q3 ≦ x ≦ p6), the elements of the first cluster P When counting the number, R1 is 3, R2 is 0, R3 is 1, R4 is 2. Therefore, R1 (p1 ≦ x <q1), which is the largest, becomes the lower approximation set.

  Also, as shown in FIG. 10, when the first cluster P is included in the second cluster Q (P） Q) (corresponding to “fourth relationship”), the maximum value and the minimum value of the first cluster P Value and the value of the second cluster Q included in the overlap of the first cluster P and the second cluster Q (the value of the second cluster Q included in the overlap) The number of values of the first cluster P included in the range from the smaller value to the larger value is calculated, and the first cluster P included in the range between the two adjacent values having the largest number is calculated. The set of elements is the lower approximation set. In the example shown in FIG. 10, the relationship between the maximum value (p5) and the minimum value (p1) of the first cluster P and the values (q3, q4, q5, q6) of the second cluster included in the first cluster P is , P1 <q3 <q4 <q5 <q6 <p5. Therefore, R1 (p1 ≦ x ≦ q3), R2 (q3 ≦ x ≦ q4), R3 (q4 ≦ x ≦ q5), R4 (q5 ≦ x ≦ q6), R5 (q6 ≦ x ≦ p5) , The number of elements of the first cluster P is counted as 1 for R1, 0 for R2, 1 for R3, 2 for R4, and 1 for R5. Therefore, R4 (q5 <x <q6), which is the largest, becomes the lower approximation set. Here, in the range specified by two values, if the value is an element of the first cluster P, an equal sign is included (≧, ≦), and if the value is an element of the second cluster Q, the equal sign is included. Not included (>, <).

Next, the effectiveness evaluation value will be described. This validity evaluation value is a value for evaluating the validity of the predicate logical expression created from the combination of the condition attributes. The effectiveness evaluation value is given by Equation 6. Note that ρ is a parameter indicating the degree to which the number of condition attributes is reflected in the evaluation, and is changed depending on the situation where information classification is performed. Usually, it is set to about 2. The lower approximate event data number is the number of event data whose condition attribute value is included in the lower approximate set, and the upper approximate event data number is the number of event data whose condition attribute value is included in the upper approximate set. The effectiveness evaluation value increases as the value increases, and the effectiveness decreases as the value decreases. That is, a predicate logical expression that is less effective is more likely to be unable to be accurately classified even if event data is classified based on the predicate logical expression. Therefore, predicate logical expressions below a predetermined value are not adopted.

  In ordinary rough set theory, approximate accuracy is used as a value to evaluate the accuracy, and if the decision attribute can be completely classified by the combination of the condition attribute, the upper approximation set and the lower approximation set are the same set. The approximation accuracy is “1”, which is evaluated as high accuracy. In addition, as the number of condition attributes constituting a combination of condition attributes increases, the upper approximation set and the lower approximation set become closer, so that the accuracy tends to increase although the number of condition attributes is simply large. Therefore, in the present invention, the effectiveness evaluation value is obtained by weighting according to the number of condition attributes and the number of event data.

  As described above, a predicate logical expression for determining a determination attribute of event data can be created using a lower approximation set of rough set theory. In addition, by calculating the validity evaluation value, the validity of the created predicate logical expression can also be evaluated, so that a predicate logical expression having higher effectiveness can be preferentially used.

  Next, as a specific example of the present invention, predicate logic for translating transaction data into accounting credit or debit account items, with “events” as accounting transactions and “event data” as transaction data generated in accounting transactions. Explains the creation of formulas and journal entries for transaction data.

  First, with reference to FIG. 11, the flow of processing of accounting journal entry rules and accounting journal entries will be described. FIG. 11 is a diagram showing the relationship between data and processing in this embodiment. In this accounting entry rule, an accounting entry rule that incorporates the characteristics of the company is created for each company, rather than creating an accounting entry rule (predicate formula) that applies to any company. As shown in FIG. 11, in the accounting journal rule creation process 30, a predicate logical expression 34 is created by adding a common rule 32 in accounting transactions and a company characteristic rule 33 unique to the company to the past journal result 31. . In this case, the determination attributes are a debit account item and a credit account item. In the journal processing 36, the transaction data 35 is journalized based on the predicate logical formula 34, the debit account item and the credit account item are determined, and the journal result 37 is created.

  The common rule 32 is derived based on information of an accountant (accounting person, accountant, tax accountant, etc.), for example, ““ Date does not affect the determination of account items ”. The rule is “delete the attribute of“ date ”from the journal entry result”. The company characteristic rule 33 is “if the“ partner ”is“ M research company ”” “summary” is “information provision fee” withdrawal transaction, there is “information provision contract” and “debit account item” "Is a" sales commission "" derived from information such as "If the" partner "is a" M research company "" summary "is a" information provision fee "withdrawal transaction, The rule is that “Yes” is set in the “Provision contract” attribute, and “No” is set for withdrawal transactions where “Description” is not “M research company” and “Information provision fee”. A new “information provision contract” attribute is added to the past journal result.

  These rules are described in a format that can be deciphered by the accounting journal rule creation program. Also, company information such as the number of employees and annual sales is input here. For example, the rule “delete the attribute of“ date ”from journal results” is expressed as “journal (debit account, credit account):-date (X1), fail.”. Expressed as number (56), annual sales of 25 million yen is expressed as annual sales (25000000).

  Next, the computer 100 that performs accounting journal rule creation and accounting journal processing will be described with reference to FIG. FIG. 12 is a block diagram showing an electrical configuration of the computer 100. As shown in FIG. 12, the computer 100 is provided with a CPU 110 that controls the computer 100. The CPU 110 temporarily stores various types of data, a RAM 101 that stores various data, a ROM 102 that stores BIOS, and the like. An I / O interface 103 that performs mediation is connected. A hard disk device 120 is connected to the I / O interface 103, and the past journal result storage area 121, common rule storage area 122, company characteristic rule storage area 123, and predicate logical expression storage are stored in the hard disk device 120. An area 124, a transaction data storage area 125, a journal result storage area 126, a program storage area 127, and other information storage areas 128 are provided.

  The past journal result storage area 121 stores past journal results for generating a predicate logical expression, and the common rule storage area 122 stores common rules 32 for general accounting processing. The company characteristic rule storage area 123 stores a company characteristic rule 33 unique to the company. The predicate logical expression storage area 124 stores the predicate logical expression created by the accounting journal entry rule creation process. The transaction data storage area 125 stores transaction data to be journalized in the accounting journal processing. The journal result storage area 126 stores the result of journalizing the transaction data stored in the transaction data storage area 125 in the accounting journal processing. The program storage area 127 stores a program executed by the CPU 110, and the other information storage area 128 stores other information used by the computer 100.

  In addition, a video controller 104, a key controller 105, a CD-ROM drive 106, an FD drive 107, and a mouse 108 are connected to the I / O interface 103, and a display 114 is connected to the video controller 104. A keyboard 115 is connected to the key controller 105, and a control program for the computer 100 is stored in a CD-ROM 116 inserted into the CD-ROM drive 106, and the hard disk device 120 is loaded from the CD-ROM 116 at the time of introduction. Are stored in the program storage area 127. Further, the FD drive 107 can read and write to the flexible disk 117.

  The past journal result 31, the common rule 32, and the company characteristic rule 33 may be directly input to the computer 100 by the keyboard 115 or the mouse 108, or may be input by a storage medium such as the CD-ROM 116 or the flexible disk 117. good. Alternatively, the computer 100 may be connected to a LAN or the Internet and input via a network.

  Next, the details of the accounting journal processing will be described with reference to FIGS. First, with reference to FIG.16 and FIG.17, the determination table used as the basic data in an accounting journal rule preparation process is demonstrated. The flow of the accounting journal entry rule creation process will be described with reference to the flowcharts of FIGS. In this case, a case of a combination of condition attributes (only “summary”) will be described with reference to FIGS. 18 to 22 as a specific example. Subsequently, the case of a combination of condition attributes (only “current account withdrawal”) will be described with reference to FIGS. 23 to 27, and the combination of condition attributes (“ The case of "other party", "information provision contract") will be described as a specific example.

  FIG. 13 is a flowchart of accounting journal rule creation processing, FIG. 14 is a flowchart of attribute combination processing of accounting journal rule creation processing, and FIG. 15 is a flowchart of approximate set creation processing under attribute combination processing. 16 is a schematic diagram of the decision table 501 created based on the past journal result 31 stored in the past journal result storage area 121. FIG. 17 includes the common rule 32 and the company characteristic rule 33. It is a schematic diagram of the determination table 502 after having performed. 18 is a schematic diagram of the partial determination table 201 for the condition attribute “explanation”, FIG. 19 is a schematic diagram of the partial determination table 202 for the condition attribute “explanation”, and FIG. It is a schematic diagram of the partial determination table 203 for "." The decision tables 501 and 502 and the partial decision tables 201 to 203 are intermediate data created and used in the RAM 101 by the accounting journal rule creating program. FIG. 21 is a schematic diagram of the cluster information storage area 301 of the RAM 101 that stores information on the first cluster and the second cluster regarding the condition attribute “explanation”, and FIG. 22 shows information on the set regarding the condition attribute “explanation”. 2 is a schematic diagram of a predicate formula information storage area 302 of a RAM 101 that stores a predicate formula.

  First, with reference to FIG.16 and FIG.17, the determination table used as the basic data in an accounting journal rule preparation process is demonstrated. As shown in FIG. 16, the decision table 501 is composed of a plurality of transaction data in a table format in which each attribute of transaction data is divided into a condition attribute and a decision attribute, and each row has individual data and a column has attributes. Has been. In the example shown in FIG. 16, “date”, “description”, “partner”, “cash deposit”, “cash withdrawal”, “current account deposit”, “current account withdrawal”, “ordinary deposit”, “Normal deposit withdrawal” and “Target employee ratio (%)” are condition attributes, and “Debit account item” and “Credit account item” are decision attributes. And it is comprised by the data of 30 past journal results.

  In the example shown in FIG. 16, in the first journal result data, the transaction data whose “summary” is “today's sales” and “cash deposit” is “86080” is “cash” for “debit account” is “credit”. The account is journalized in “Sales”. In the second journal result data, the transaction data for “Description” is “Purchase”, “Partner” is “A Tofu Shop”, and “Cash Withdrawal” is “37972” is “Debit Account”. “Product purchases” and credit account items are journaled as “cash”. In the third journal entry result data, the transaction data for “Summary” is “Social Insurance Premium”, “Partner” is “T Social Insurance Office”, and “Payment with Deposits” is “129204” on 4/2. “Account” is journalized as “statutory welfare expenses”, and credit account is “regular deposit”. In the fourth journal result data, 4/3, “Summary” is “Today's sales”, “Cash deposit” is “136650”, Transaction data is “Debit account” is “Cash”, and Credit account is “Sales” Journalized in In the fifth journal entry result data, the transaction data with “Description” set to “Purchase”, “Partner” set to “A Tofu Store”, and “Cash Withdrawal” set to “9713” is set to “Debit Account”. “Product purchases” and credit account items are journaled as “cash”. In the 6th journal entry result data, the transaction data with “Summary” is “Purchase”, “Partner” is “K liquor”, “Cash withdrawal” is “3675”, and “Debit account” is “Product” “Purchase” and credit account items are journaled as “cash”. In the 7th journal entry result data, the transaction data for “Summary” is “Whiteboard Marker”, “Partner” is “S Office Supplies Store”, and “Cash Withdrawal” is “850” on 4/3. "" Equipment and consumables expenses ", and the credit account item is" cash ". In the 8th journal entry result data, the transaction data of “Tendee” is “T gas” and “Checking account withdrawal” is “20405” is 4/3, “Debit account” is “Utilities”, Credit account Is journalized in "Checking account". In the 9th journal result data, the transaction data for “Corporate” is “C card company”, “Current account withdrawal” is “4200”, and “Debit account” is “Equipment and consumables”. The credit account item is journalized in “Checking account”. In the 10th journal entry result data, the transaction data with “Description” as “garbage disposal ticket”, “Destination” as “K supermarket”, and “Cash withdrawal” as “1080” “Miscellaneous expenses” and credit account items are journalized as “cash”.

  In the eleventh journal entry result data, the transaction data with “summary” is “telephone charges”, “partner” is “N company”, and “current account withdrawal” is “6074” is 4/10. "Communication transportation expenses" and credit account items are journalized as "Checking account". In the twelfth journal entry result data, the transaction data with “Description” set to “Telephone fee”, “Payee” set to “TT Net”, and “Current account withdrawal” set to “1320” "Communication transportation expenses" and credit account items are journalized as "Checking account". In the 13th journal entry result data, on 4/14, “summary” is “Kaoriyo”, “partner” is “U”, and “cash withdrawal” is “10000”. "Accounting expenses" and credit account items are journalized as "Cash". In the 14th journal result data, on 4/15, the transaction data with “Summary” is “Employee Travel”, “Cash Withdrawal” is “300000”, and “Target Employee Ratio (%)” is “5”. "" Is entered as "office worker salary" and the credit account item is entered as "cash". In the 15th journal result data, on 4/15, the transaction data for “Description” is “Information provision fee”, “Partner” is “M research company”, and “Current account withdrawal” is “200000”. "Is charged as" sales commission ", and the credit account is journalized as" current account ". In the 16th journal result data, on 4/17, the transaction data for “Description” is “Purchase (March)”, “Partner” is “N store”, and “Current account withdrawal” is “478155”. “Account” is journalized as “Product purchases”, and credit account is “Current account”. In the 17th journal result data, on 4/25, “Description” is “Telephone fee”, “Partner” is “K company”, “Current account withdrawal” is “623” and the transaction data is “Debit account” "Communication transportation expenses" and credit account items are journalized as "Checking account". In the 18th journal entry result data, the transaction data with “Description” as “Advisor Fee”, “Partner” as “Y Accounting Office”, and “Current Account Withdrawal” as “47500” is “Debit Account”. "" And "administrative expenses", and credit account items are journalized as "current account". In the 19th journal entry result data, the transaction data for “Rating” is “Rent”, “Tender” is “T-Bill”, and “Current Account Withdrawal” is “300000” is “Debit Account”. The “rent for rent” and credit account items are journalized as “current account”. In the 20th journal result data, on 4/27, “Summary” is “Salary”, “Partner” is “A Senior Managing Director”, and “Cash Withdrawal” is “450,000”. "Remuneration" and credit account items are journalized as "Cash".

  In the 21st journal entry result data, the transaction data for “Summary” is “Salary”, “Partner” is “2 Employees”, and “Cash withdrawal” is “77420” on 4/27. “Clerical employee salary” and credit account are journalized as “cash”. In the 22nd journal result data, on May 22nd, “Summary” is “Information provision fee”, “Partner” is “Mr. P”, “Current account withdrawal” is “150,000” and the transaction data is “Debit account item”. Is accounted for as “accounting expense” and credit account as “current account”. In the 23rd journal result data, on May 25th, transaction data with “Description” is “Employee Travel”, “Cash Withdrawal” is “2500000”, and “Target Employee Ratio (%)” is “95”. Is credited with “welfare benefits” and credit account is “cash”. In the 24th journal result data, the transaction data with “summary” is “Tea fee (internal)”, “cash withdrawal” is “10000”, and “target employee ratio (%)” is “100” “Debit Account” is journalized as “Benefit Expenses” and Credit Account is “Cash”. In the 25th journal entry result data, the transaction data with “Summary” is “Information provision fee”, “Destination” is “Professor R at Q University”, “Current account withdrawal” is “160000” “Account” is accounted for as “accounting expense” and credit account item as “current account”. In the 26th journal entry result data, the transaction data with “Summary” is “Information provision fee”, “Partner” is “Mr. S”, “Current account withdrawal” is “120,000” will be “Debit account” on August 29th. Is accounted for as “accounting expense” and credit account as “current account”. In the 27th journal result data, the transaction data with “Description” is “Employee Travel”, “Cash Withdrawal” is “240000”, and “Target Employee Ratio (%)” is “4” on 10/15. "" Is entered as "office worker salary" and the credit account item is entered as "cash". In the 28th journal entry result data, the transaction data with “Description” is “Information provision fee”, “Partner” is “M research company”, and “Current account withdrawal” is “150,000” is 10/15. "Is charged as" sales commission ", and the credit account is journalized as" current account ". In the 29th journal entry result data, the transaction data with “summary” is “tea fee (in-house)”, “cash withdrawal” is “10000”, and “target employee ratio (%)” is “100” is 12/1. “Debit Account” is journalized as “Benefit Expenses” and Credit Account is “Cash”. In the 30th journal result data, on 12/25, “description” is “employee trip”, “cash withdrawal” is “1000000”, “target employee ratio (%)” is “90” transaction data is “debit account item” Is credited with “welfare benefits” and credit account is “cash”.

  FIG. 17 is a schematic diagram of the determination table 502 after the common rule 32 and the company characteristic rule 33 are taken into consideration. In this embodiment, the common rule 32 is “delete the attribute of“ date ”from the journal entry result”, and the company characteristic rule is “the“ partner ”is“ M survey company ”“ explanation ”is“ information provision fee ”. In the case of a withdrawal transaction, “present” is set in the “information provision contract” attribute, and “none” is set for withdrawal transactions in which the “summary” is “information provision fee”. Therefore, in the decision table 502, the condition attribute “date” of the decision table 501 is deleted, an “information provision contract” attribute is added, and the 15th and 28th journal result data is “summary” is “information provision fee”, Since “Destination” is “M Survey Company”, “present” is set in the “information provision contract” attribute, and “summary” is “information provision fee” in the 22nd, 25th and 26th journal result data. However, since “the other party” is not “M research company”, “none” is set in the “information provision contract” attribute.

  Next, the accounting journal entry rule creation process will be described with reference to the flowchart of FIG. First, the past journal result 31 stored in the past journal result storage area 121 is read (S51), and a decision table 501 (see FIG. 16) is created and stored in the RAM 101 (S52). The common rule 32 is read from the common rule storage area 122 (S53), the decision table 501 is modified based on the common rule 32 (S54), and then the company characteristic rule 33 is read from the company characteristic rule storage area 123. (S55) The decision table 501 is modified based on the company characteristic rule 33 (S56). The determination table after correction in S54 and S56 is a determination table 502 (see FIG. 17).

  Then, all combinations of condition attributes are created (S57). The combinations created here are stored in the RAM 101. In the decision table 502 shown in FIG. 17, (summary), (destination), (cash deposit), ..., (information provision contract), (summary, destination), (summary, cash deposit), ..., (summary) , Counterparty, cash deposit), ..., (savings deposit, savings deposit withdrawal, information provision contract), ..., (explanation, counterparty, cash deposit, cash withdrawal, current account deposit, current account withdrawal), All the conditions such as (summary, counterparty, cash deposit, cash withdrawal, current account deposit, current account deposit, savings deposit, ordinary deposit withdrawal, target employee ratio (%), information provision contract) A combination of attributes is created.

  Then, the first combination is designated (S58). Since all the combinations have not been processed yet (S59: NO), the condition attribute combination processing is performed for the designated combination (S60, see FIG. 14). For example, first, a combination of only the condition attribute “summary” is designated, a condition attribute combination process is performed, a predicate logical expression is created, and stored in the predicate logical expression information storage area of the RAM 101 (S60). Then, the process returns to S58, and the next combination of condition attributes (for example, the combination of “other party” only) is designated (S58), and all the combination processes have not been completed (S59: NO). An attribute combination process is performed (S60), and the process returns to S58. When such processing is repeated and the conditional attribute combination processing is completed for all combinations, and the predicate logical expression is created and stored (S59: YES), the process proceeds to S61.

  Here, the condition attribute combination process will be described with reference to the flowchart of FIG. 14 and FIGS. 18 to 22, with a specific example of creating a predicate logical expression from a combination of condition attributes (only “summary”).

  First, the partial decision table 201 including only the condition attribute and the decision attribute constituting the combination of the condition attribute designated in S59 is created and stored (S71). This partial determination table 201 is obtained by replacing the columns of the determination table 502 with only the condition attribute and the determination attribute constituting the combination (see FIG. 18). In the combination (summary), as shown in FIG. 18, the partial attribute table 201 is composed of three columns of only the condition attribute “summary” and the decision attributes “debit account item” and “credit account item”.

  Then, the data (column) having no value in the condition attribute “summary” is deleted from the partial determination table 201 (S72). The partial determination table 202 after deletion is a partial determination table 202 shown in FIG. In the example shown in FIG. 19, data of “debit account” in the eighth column of the partial determination table 201 is “utility expenses”, “credit account” is “current account”, and “debit account” in the ninth column. The data of “equipment / consumables expenses” and “credit account item” of “current account” are deleted, and a partial determination table 202 is obtained.

  Next, the partial determination table 202 is sorted by the value of the determination attribute (S73), and the first cluster and the second cluster are determined (S74). The partial determination table 203 shown in FIG. 20 is a partial determination table obtained by sorting the partial determination table 202 and numbering combinations of determination attributes. The first cluster is a set of data having the same decision attribute value, and is created for each combination of decision attribute values in the partial decision table 202. And the data which has the value of a decision attribute other than that with respect to a 1st cluster are classified into a 2nd cluster. Therefore, the second cluster is a complement of the first cluster in the data belonging to the partial decision table 202.

  In the partial determination table 203 shown in FIG. 20, the first first cluster includes two items whose debit account item is “cash”, the credit account item is “sales”, and the condition attribute “summary” is “sales today”. The data belongs, and the second first cluster has one data whose debit account item is “Miscellaneous expenses”, the credit account item is “Cash”, and the condition attribute “Description” is “Waste disposal ticket”. In the third cluster, the debit account item is “Clerical employee salary”, the credit account item is “Cash”, and the condition attribute “Description” is “Employee travel” and “Salary”. ing. And the 4th 1st cluster has 3 items with debit account item “Product purchases”, credit account item “Cash” and condition attribute “Description” “Purchase”. The first cluster of 5 belongs to one data in which the debit account item is “accounting expense”, the credit account item is “cash”, and the condition attribute “summary” is “Kaoriyo”. The cluster includes one piece of data in which the debit account item is “equipment / consumables expense”, the credit account item is “cash”, and the condition attribute “explanation” is “whiteboard marker”.

  In the first cluster, the debit account item is “benefits expenses”, the credit account item is “cash”, and the condition attributes “summary” are “employee travel” and “tea fee (in-house)” 4 The 8th 1st cluster has one data whose debit account item is “Executive Remuneration”, the credit account item is “Cash”, and the condition attribute “Summary” is “Salary”. In the ninth first cluster, one item with the debit account item “management expenses”, the credit account item “current account”, and the condition attribute “explanation” “advisory fee” belongs. The tenth first cluster has three data with the debit account item “accounting expense”, the credit account item “current account”, and the condition attribute “description” “information provision fee”. The eleventh first cluster belongs to one data item whose debit account item is “Product purchases”, the credit account item is “Checking account”, and the condition attribute “Description” is “Purchase (March)” In the twelfth first cluster, one piece of data having the debit account item “rental rent”, the credit account item “current account”, and the condition attribute “explanation” “rent” belongs. The thirteenth first cluster includes three data items with the debit account item “communication transportation expenses”, the credit account item “current account”, and the condition attribute “summary” “phone bill”. The fourteenth first cluster includes two pieces of data in which the debit account item is “sales fee”, the credit account item is “current account”, and the condition attribute “explanation” is “information provision fee”. In the first cluster, one item of data in which the debit account item is “statutory welfare expense”, the credit account item is “ordinary deposit”, and the condition attribute “explanation” is “social insurance premium” belongs.

  That is, in the partial determination table 203, 15 first clusters are generated. In S 74, the generated information about the first to fifteenth first clusters is stored in the cluster information storage area 301 of the RAM 101. As shown in FIG. 21, in the cluster information storage area 301, for each combination of determination conditions, “attribute name” and “data type” of the condition attribute, “value” of the data belonging to the first cluster, and the first condition attribute value Data relating to the set, “value” of data belonging to the second cluster, and data relating to the second condition attribute value set are stored.

  In the example shown in FIG. 21, since the data type of the condition attribute is “character string”, the “value” is set in the first cluster and the second cluster. The set of condition attribute values set in the “value” of the “first cluster” is the same as the “first condition attribute value”. Similarly, a set of condition attribute values set in “value” of “second cluster” is “second condition attribute value set”.

  In the example shown in FIG. 21, in the first combination (“cash”, “sales”), the condition attribute is “summary” with the data type “character string”, and the number of data belonging to the first cluster is 1. Therefore, the condition attribute value is “Today's sales”, the number of data belonging to the second cluster is 12, and the condition attribute values are “Tea fee (in-house)”, “Salary”, “Kanen”, “Garbage” “Processing ticket”, “Advisor fee”, “Purchase”, “Employee travel”, “Social insurance fee”, “Information provision fee”, “Telephone fee”, “Whiteboard marker”, “Rent”. Further, in the second combination (“miscellaneous expenses”, “cash”), the condition attribute is “summary” with the data type “character string”, and the number of data belonging to the first cluster is one. The value is “garbage disposal ticket”, the number of data belonging to the second cluster is 12, and the condition attribute values are “tea fee (in-house)”, “salary”, “salary fee”, “advisory fee”, “purchase” ”,“ Employee Travel ”,“ Social Insurance Fee ”,“ Information Provision Fee ”,“ Telephone Fee ”,“ Whiteboard Marker ”,“ Sales Today ”, and“ Rent ”. In addition, the third to sixth combinations are omitted, and in the seventh combination (“benefit expenses”, “cash”), the condition attribute is “explanation” with the data type “character string”. The number of data belonging to one cluster is two, the condition attribute values are “employee trip”, “tea fee (in-house)”, the number of data belonging to the first cluster is 12, and the condition attribute value is “salary” ”,“ Kanori ”,“ Waste disposal ticket ”,“ Advisor fee ”,“ Purchasing ”,“ Employee travel ”,“ Social insurance fee ”,“ Information provision fee ”,“ Telephone fee ”,“ Whiteboard marker ”, “Today's sales” and “Rent”. In addition, the eighth to fourteenth combinations are omitted, and in the fifteenth combination (“statutory welfare expenses”, “ordinary deposit”), the condition attribute is “explanation” whose data type is “character string”. The number of data belonging to the first cluster is one, the condition attribute value is “social insurance premium”, the number of data belonging to the second cluster is 12, and the condition attribute value is “tea fee (in-house)” , “Salary”, “Kanoriyo”, “Waste disposal ticket”, “Advisor fee”, “Purchase”, “Employee travel”, “Information provision fee”, “Telephone fee”, “Whiteboard marker”, “Today's sales” "Rent".

  When the data type of the condition attribute is “number”, “maximum value” and “minimum value” of “first cluster”, “value” of data, and “maximum value” of “second cluster” ”And“ minimum value ”and the“ value ”of the data are stored (see FIG. 26). Here, the “first condition attribute value set” is a collection of numbers (an integer if the data type is an integer, a real number if the data type is an integer) that is greater than or equal to the “minimum value” and less than the “maximum value” of the “first cluster”. In addition, the “second condition attribute value set” is a collection of numbers from the “minimum value” to the “maximum value” of the “second cluster” (an integer if the data type is an integer, and a real number if the data type is a real number). The case where the data type of the condition attribute is “number” will be described in detail when the condition attribute “current account withdrawal” is described as an example. If the combination of condition attributes is composed of a plurality of condition attribute values, a first condition attribute value set and a second condition attribute value set are determined for each condition attribute value (see FIG. 32). This will also be described in detail when a combination of the condition attributes “destination” and “information provision contract” is taken as an example.

  As described above, the first cluster, the second cluster, the first condition attribute value set, and the second condition attribute value set are generated for each combination of determined attributes, and information related thereto is stored in the cluster information storage area 301. (S74) A predicate logical expression is created for each combination of decision attributes and stored in the predicate logical expression information storage area 302 (S77 to S80). Therefore, first, a combination of the first determination attributes is designated (S75). In the first combination of decision attributes, since all the combinations of decision attributes have not been finished yet (S76: NO), the upper approximation set is created without finishing the processing, and the predicate logical expression information storage area 302 has Stored (S77). Information about the approximate set is stored in the predicate logical expression information storage area 302 (see FIG. 22) of the RAM 101.

  As shown in FIG. 22, in the predicate logical expression information storage area 302, various information is stored for each combination of decision attributes, similarly to the cluster information storage area 301. “Value” of the condition attribute of the data constituting the upper approximate set is set in “value” in “upper approximate set”. Since the upper approximate set is a set (first condition attribute value set) having “values” of data belonging to the first cluster as elements, the “value” of the first cluster in the cluster information storage area 301 is set. In the example shown in FIG. 22, in the first combination of decision attributes (cash, sales), a set having [Today's sales] as an element is an upper approximate set.

  When the upper approximate set is created (S77), the lower approximate set creation process is performed (S78, see FIG. 15). In the lower approximation set creation process, a lower approximation set is created based on the relationship between the first cluster and the second cluster for the combination of the determined decision attributes, and stored in the predicate logical expression information storage area 302. Here, the lower approximation set creation process will be described with reference to the flowchart of FIG.

  First, the relationship between the first condition attribute value set and the second condition attribute value set is determined for the specified combination of decision attributes (S101). This relationship is determined by the relationship between the “value” (first condition attribute value set) of the first cluster in the cluster information storage area 301 and the “value” (second condition attribute value set) of the second cluster. If there is no overlap between the first condition attribute value set and the second condition attribute value set and they are independent from each other, a “first relationship” is assumed, and there is an overlap between the first condition attribute value set and the second condition attribute value set. If one is not included in the other, a “second relationship” is established, the first condition attribute value set and the second condition attribute value set overlap, and the second condition attribute value set is the first condition attribute value. When included in the set, a “third relationship” is established, the first condition attribute value set and the second condition attribute value set overlap, and the first condition attribute value set is included in the second condition attribute value set. In this case, the “fourth relationship” is set. The “relation” in the predicate formula information storage area 302 is “1” for the first relationship, “2” for the second relationship, “3” for the third relationship, and “3” for the fourth relationship. “4” is set. If there is an overlap, the element is set in the overlap “value” in the predicate logical expression information storage area 302.

  In the example shown in FIG. 21 and FIG. 22, if the combination of designated decision attributes is the first decision attribute combination (cash, sales), it is an element of the first condition attribute value set [sales today] Does not exist in the elements of the second condition attribute value set, there is no overlap, nothing is set in the “value” of overlap in the predicate logical expression information storage area 302, and “1” is set in “relation” . Further, if the combination of designated decision attributes is the seventh decision attribute combination (benefit expenses, cash), [employee trip] is in the first condition attribute value set and the second condition attribute value set. Therefore, “employee travel” is set in the “value” of the overlap of the predicate logical formula information storage area 302, and “2” is set in “relation”.

  When the relationship between the first condition attribute value set and the second condition attribute value set is determined for the specified combination of decision attributes (S101), a lower approximate set is created according to the relationship, and the condition attributes to be configured Is set to the “value” of the “lower approximation set” in the predicate logical expression information storage area 302 (S102 to S127).

  When the relationship is the first relationship (S102: YES), the first condition attribute value set is directly used as the lower approximation set, so that the “value” of the first cluster in the cluster information storage area 301 is the predicate logical expression information. It is set to the “value” of the “lower approximation set” in the storage area 302 (S105). In the example shown in FIG. 21 and FIG. 22, if the combination of designated decision attributes is the first decision attribute combination (cash, sales), a set having [Today's sales] as an element is a lower approximation set. Has been.

  Further, when the relationship is the second relationship (S102: NO, S103: YES), the set obtained by removing the overlap from the first condition attribute value set is set as the lower approximate set. Elements obtained by subtracting the “value” of “overlap” in the predicate logical expression information storage area 302 from the “value” of one cluster are set (S106). In the example shown in FIGS. 21 and 22, if the combination of designated decision attributes is the seventh decision condition combination (benefits expenses, cash), [Employee Travel, Tea Fee (In-house)] to [Employees] [Tea fee (in-house)] excluding [Travel] is set to “value”.

  When the relationship is the third relationship (S102: NO, S103: NO, S104: YES), the processing differs depending on whether the condition attribute type is a character string type or a numeric type. Therefore, it is first determined whether or not the condition attribute type is a character string type (S110). In the case of the character string type (S110: YES), the set obtained by removing the overlap from the first condition attribute value set is set as the lower approximate set, and therefore, from the “value” of the first cluster in the cluster information storage area 301. Elements excluding “value” of “overlap” in the predicate logical expression information storage area 302 are set (S117). The processing for the case of not being a character string type (S110: NO) will be described in detail when an example of a condition attribute whose condition attribute type is “number” is described.

  Even when the relationship is the fourth relationship (S102: NO, S103: NO, S104: NO), the processing differs depending on whether the type of the condition attribute is a character string type or a numeric type. Therefore, it is first determined whether or not the condition attribute type is a character string type (S120). In the case of the character string type (S120: YES), no predicate logical expression is created, and therefore no lower approximation set is created (S127). Therefore, a predetermined value (for example, “fail”) indicating that a predicate logical expression is not created is set in the “lower approximate set” or “predicate logical expression” in the predicate logical expression information storage area 302 (S127). The processing for the case of not being a character string type (S120: NO) will be described in detail when an example of a condition attribute whose condition attribute type is “number” is described.

  As described above, when the lower approximation set is created (S78), the effectiveness evaluation value is calculated (S79). This effectiveness evaluation value is calculated by the equation shown in Equation 6. Here, ρ = 2 and the third decimal place are regarded as effective digits and rounded off. In the first combination of decision attributes shown in FIG. 22 (cash, sales), the number of event data having elements belonging to the upper approximate set is “2”, and the number of event data having elements belonging to the lower approximate set is “ 2 ”, the number of condition attributes constituting the combination of condition attributes is“ 1 ”, and the total number of event data is“ 30 ”. Therefore, the approximation accuracy is the number of event data having elements belonging to the upper approximation set. By dividing the number of event data having elements belonging to “2 ÷ 2 = 1”. Then, since the square root of the total condition attribute number “1” is “1.000”, the corrected approximate accuracy is “1 ÷ 1.000 = 1.000” by dividing the approximation accuracy “1”. Then, the effectiveness evaluation value is obtained by multiplying the approximation accuracy “1.000” by the number of event data having key points belonging to the lower approximation set, divided by the total number of event data, and “2 ÷ 30 × 1.000 = 0. .067 ".

  When the validity evaluation value is calculated and stored in the predicate logical expression information storage area 302 (S79), a predicate logical expression is created and stored in the predicate logical expression column of the predicate logical expression information storage area 302 of the RAM 101 (S80). ). Specifically, the left side of the predicate logical expression is “journal (debit account item name, credit account item name)”, and the character string in which the combination of the decision attributes is set in parentheses is stored. The “journal” may be an arbitrary character string, and in this embodiment, the “journal” is used because a predicate logical expression for journalizing transaction data is created. On the right side, “condition attribute 1 (A), member (A, [condition attribute value, condition attribute value,...]), Condition attribute 2 (B), member (B, [condition attribute value, condition Attribute value ...])) ... ". In “condition attribute 1, condition attribute 2,...”, A condition attribute name is set according to a combination of designated condition attributes. “A” and “B” are variables used when substituting transaction data into the predicate logical expression. In addition, in () of “member (A, [condition attribute value, condition attribute value,...])”, A list of values belonging to the first condition attribute value set is listed in addition to the variable A. .

  In the example of the first determination attribute combination shown in FIG. 22, the determination attribute is (cash, sales), the condition attribute is “summary” only, and the element of the first condition attribute value set is “sales today”. Therefore, the predicate logical expression becomes “journal (cash, sales):-summary (A), member (A, [sales today]).”. This is a predicate logical expression when the value of the condition attribute is a character string type. The predicate logical expression in the case of not being a character string type will be described in detail when explaining an example of a condition attribute whose condition attribute type is “number”. Further, when a predetermined value indicating that a predicate logical expression is not created is set in the “lower approximation set” or “predicate logical expression”, the predicate logical expression is not created.

  In this way, when the predicate logical expression is created and stored for the specified combination of decision attributes (S77 to S80), the process returns to S75, and the next combination of decision attributes is designated (S75). A predicate formula for the combination is created and stored (S77-S80). If the processes of S75 to S80 are repeatedly performed for all combinations of determined attributes (S76: YES), the condition attribute combination process is terminated and the process returns to the predicate logical expression creation process (see FIG. 13). Then, the process proceeds to S58, and processing for the next combination of condition attributes is performed. When the condition attribute combination processing (S60) is completed for all the condition attribute combinations, and predicate logical expressions are created and stored for all the combinations of the decision attributes for all the condition attribute combinations, respectively (S59: YES), the process proceeds to S61.

  In the example of other combinations of decision attributes shown in FIG. 22, the second decision attribute combination (miscellaneous expenses, sales) has only the “requirement” condition attribute, and the element of the first condition attribute value set is “garbage”. The number of event data having elements belonging to the upper approximation set is “1”, the number of event data having elements belonging to the lower approximation set is “1”, and the number of condition attributes constituting the combination of condition attributes Is “1” and the total number of event data is “30”. Therefore, the approximation accuracy is obtained by dividing the number of event data having elements belonging to the lower approximation set by the number of event data having elements belonging to the upper approximation set. 1 ÷ 1 = 1 ”. Then, since the square root of the total condition attribute number “1” is “1.000”, the corrected approximate accuracy is “1 ÷ 1.000 = 1.000” by dividing the approximation accuracy “1”. Then, the effectiveness evaluation value is obtained by multiplying the approximate accuracy “1.000” by the number of event data having the key belonging to the lower approximate set divided by the total number of event data to obtain “1 ÷ 30 × 1.000 = 0. .033 ". The predicate logical expression becomes “journal (cash, sales):-summary (A), member (A, [garbage disposal ticket]).”.

  In the seventh combination of decision attributes (benefit expenses, cash), the condition attribute is “Summary” only, and the element of the first condition attribute value set is “employee trip, tea fee (in-house)”. The number of event data having elements belonging to the approximate set is “6”, the number of event data having elements belonging to the lower approximate set is “2”, the number of condition attributes constituting the combination of condition attributes is “1”, and all event data Since the number is “30”, the approximation accuracy is “2 ÷ 6 = 0.333” by dividing the number of event data having elements belonging to the lower approximation set by the number of event data having elements belonging to the upper approximation set. It becomes. Then, since the square root of the number of all condition attributes “1” is “1.000”, the corrected approximate accuracy is “0.333 ÷ 1.000 = 0.333” by dividing the approximate accuracy “0.333”. It becomes. Then, the effectiveness evaluation value is obtained by multiplying the approximate accuracy “0.333” by the number of event data having the key belonging to the lower approximate set divided by the total number of event data, and “2 ÷ 30 × 0.333 = 0. .022 ". The predicate logical expression becomes “Journal (benefit expenses, cash):-Summary (A), member (A, [tea fee (in-house)])”.

  In addition, since the combination of the fifteenth decision attributes (statutory welfare expenses, savings deposit) is only “summary” as the condition attribute and “social insurance premium” as the element of the first condition attribute value set, it belongs to the upper approximate set. The number of event data having elements is “1”, the number of event data having elements belonging to the lower approximation set is “1”, the number of condition attributes constituting the combination of condition attributes is “1”, and the total number of event data is “30”. Therefore, the approximation accuracy is “1 ÷ 1 = 1” by dividing the number of event data having elements belonging to the upper approximation set by the number of event data having elements belonging to the lower approximation set. Then, since the square root of the total condition attribute number “1” is “1.000”, the corrected approximate accuracy is “1 ÷ 1.000 = 1.000” by dividing the approximation accuracy “1”. Then, the effectiveness evaluation value is obtained by multiplying the approximate accuracy “1.000” by the number of event data having the key belonging to the lower approximate set divided by the total number of event data to obtain “1 ÷ 30 × 1.000 = 0. .033 ". The predicate logical expression becomes “Journal (legal welfare expenses, ordinary deposit):-Summary (A), member (A, [social insurance premium]).”.

  Then, in S61 to S64 of the accounting journal entry rule creation processing of FIG. 13, the predicate logical expressions stored in the predicate logical expression information storage area 302 of the RAM 101 are sequentially read (S61), and the processing for all the predicate logical expressions is performed. If not completed (S62: NO), it is determined whether or not the validity evaluation value of the predicate logical expression is equal to or greater than a predetermined reference value (for example, 0.01) (S63). If it is equal to or greater than the predetermined reference value (S63: YES), the predicate logical expression is highly effective and is stored in the predicate logical expression storage area 124 of the hard disk device 120 together with the validity evaluation value (S64). Then, returning to S61, the next predicate logical expression is read (S61). If the validity evaluation value is less than or equal to the predetermined reference value (S63: NO), the predicate logical expression is not adopted because it has low validity, and is not stored in the predicate logical expression storage area 124. Then, the next predicate logical expression is read (S61). When the evaluation of the validity is completed for all the predicate logical expressions by repeating the processes of S61 to S64 (S62: YES), the predicate logical expressions stored in the predicate logical expression storage area 124 are converted into the validity evaluation values. Sort in descending order (S65). Then, the accounting journal entry rule creation process ends.

  Thus, the created predicate logical expressions are stored in the predicate logical expression storage area 124 in descending order of their effectiveness evaluation values.

  Next, as an example when the value of the condition attribute is “numerical type”, the condition attribute combination in the case where the combination of the condition attributes specified in S59 of the predicate formula creation process is “current account withdrawal” only. The processing (see FIG. 14) will be described with reference to FIGS.

  FIG. 23 is a schematic diagram of the partial determination table 211 for the condition attribute “current account withdrawal”, and FIG. 24 is a schematic diagram of the partial determination table 212 for the condition attribute “current account withdrawal”. FIG. 26 is a schematic diagram of the partial determination table 213 for the condition attribute “current account withdrawal”, and FIG. 26 is a diagram of the RAM 101 that stores information on the first cluster and the second cluster for the condition attribute “current account withdrawal”. FIG. 27 is a schematic diagram of the cluster information storage area 311. FIG. 27 is a schematic diagram of the predicate logical formula information storage area 312 of the RAM 101 that stores set information and predicate logical formulas for the condition attribute “current account withdrawal”.

  First, the partial determination table 211 (see FIG. 23) having only the condition attribute “current account withdrawal” is created (S71). Then, as shown in FIG. 24, the data (column) having no value for the condition attribute “summary” is deleted from the partial determination table 211 (S72). In the example shown in FIG. 24, the first to seventh columns, the tenth column, the thirteenth column, the fourteenth column, the twenty-first column, the twenty-third column, the twenty-third column, the twenty-seventh column, the 29th column, and the thirty column of the partial determination table 201. The data is deleted and the partial determination table 212 is obtained.

  Next, the partial determination table 212 is sorted by the value of the determination attribute (S73), and the first cluster and the second cluster are determined (S74). The partial determination table 213 shown in FIG. 25 is a partial determination table in which the partial determination table 212 is sorted and the combination of determination attributes is numbered.

  In the partial determination table 213 shown in FIG. 25, in the first first cluster, the debit account item is “administrative expenses”, the credit account item is “current account”, and the amount of the condition attribute “current account withdrawal” is “47500”. ”Belongs to the second cluster, the debit account item is“ accounting expense ”, the credit account item is“ current account ”, and the amount of the condition attribute“ current account withdrawal ”is“ Three data of “120,000”, “150,000”, and “160000” belong, and the third first cluster has a debit account item of “product purchase” and a credit account item of “current account”. A single attribute with the attribute “Checking withdrawn” is “478155”, and the fourth first cluster has a debit account of “Utilities” and a credit account of “Checking”. , Condition attribute “Checking The amount of "one of the data is" 20405 "belongs. The fifth first cluster has one data whose debit account item is “rental rent”, credit account item is “current account”, and the condition attribute “current account withdrawal” is “300000”. In the sixth cluster, the debit account item is “communication transportation expenses”, the credit account item is “current account”, and the condition attribute “current account withdrawal” is “623”, “1320”, Three pieces of data “6074” belong to the seventh cluster. The debit account item is “sales fee”, the credit account item is “current account”, and the amount of the condition attribute “current account withdrawal” 2 data of which “150,000” and “200000” belong, the eighth cluster has the debit account item “equipment / consumables expenses”, the credit account item “current account”, and the condition attribute The amount of “withdrawal account withdrawal” is One of the data is part of a 4200 ".

  In S74, the generated information about the first to eighth first clusters is stored in the cluster information storage area 311 of the RAM 101 (see FIG. 26). In the example shown in FIG. 26, since the data type of the condition attribute is “number”, the “maximum value”, “minimum value”, and “value” are set in the first cluster and the second cluster. In the “value” of the first cluster, the value of the condition attribute of the data constituting the first cluster (in this case “current account withdrawal”) is set, and in the “value” of the second cluster, the value of the first cluster is set. Is set to the value of the condition attribute (“current account withdrawal”) of the data that constitutes. Here, a set of numbers included in the range from the “minimum value” to the “maximum value” of the first cluster (an integer if the data type is an integer, a real number if the data type is a real number) is a “first condition attribute value set”. The set of numbers included in the range from the “minimum value” to the “maximum value” of the second cluster (integer if the data type is integer, real number if the data type is real) is “second condition attribute value set” Is equivalent to.

  In the example shown in FIG. 26, in the third combination (“product purchase amount”, “current account”), the condition attribute is “current account withdrawal” whose data type is “integer”, and the first condition attribute value The range of the set (element is X) is X = 478155, the condition attribute value in the first cluster (current account withdrawal amount) is [478155], and the range of the second condition attribute value set (element is Y) ) Is 623 ≦ Y ≦ 300,000, and the condition attribute value (the amount of current account withdrawal) in the second cluster is [623, 1320, 4200, 6074, 20405, 47500, 120,000, 150000, 150000, 160000, 200000, 300,000] It is said that.

  In the sixth combination (“communication transportation cost”, “current account”), the condition attribute is “current account withdrawal” whose data type is “integer”, and the range of the first condition attribute value set is 623 ≦ When X ≦ 6074, the condition attribute value (current account withdrawal amount) in the first cluster is [623, 1320, 6074], the range of the second condition attribute value set is 4200 ≦ Y ≦ 478155, and the condition in the second cluster The attribute value (amount of current account withdrawal) is [4200,20405,47500,120,000,150,000,150,000,160000,200000,300,000,478155]. In the seventh combination (“sales commission”, “current account”), the condition attribute is “current account withdrawal” with the data type “integer”, and the range of the first condition attribute value set is 150,000 ≦ X ≦ 200000, the condition attribute value in the first cluster (current account withdrawal amount) is [150,000,200000], the range of the second condition attribute value set is 623 ≦ Y ≦ 478155, and the condition attribute value in the second cluster ( The amount of the current account withdrawal) is [623, 1320, 4200, 6074, 20405, 47500, 120,000, 150000, 160000, 300000, 478155]. The first combination, the second combination, the fourth combination, the fifth combination, and the eighth combination are omitted.

  As described above, when the first cluster and the second cluster are generated for each combination of decision attributes, and information about the first cluster and the second cluster are stored in the cluster information storage area 311 (S74), the predicate logic is set for each decision attribute combination. An expression is created and stored in the predicate logical expression column of the predicate logical expression information storage area 312 (S77 to S80). First, the first combination of decision attributes is designated (S75). In the first combination of decision attributes, since all combinations of decision attributes have not yet been completed (S76: NO), an upper approximate set is created (S77). Information about the approximate set is stored in the predicate logical expression information storage area 312 (see FIG. 27) of the RAM 101.

  As shown in FIG. 27, in the predicate logical expression information storage area 312, various information is stored for each combination of decision attributes, as in the cluster information storage area 311. The upper approximate set is a set of numbers (the first condition attribute value set that is an integer if the data type is an integer and a real number if the data type is a real number) included in the range from the “minimum value” to the “maximum value” of the first cluster. Therefore, the “maximum value” and “minimum value” of the first cluster in the cluster information storage area 311 are set.

  In the example shown in FIG. 26 and FIG. 27, when the combination of the specified decision attributes is the third combination (“product purchase amount”, “current account”), the set in the range of X = 478155 is the top. It is an approximate set. Further, when the processes of S75 to S80 are repeated and the sixth combination (“communication transportation cost”, “current account”) is designated, the set in the range of 623 ≦ X ≦ 6074 is the upper approximate set. When the seventh combination (“sales fee”, “current account”) is designated, the set in the range of 150,000 ≦ X ≦ 200000 is set as the upper approximate set. In FIG. 27, the first combination, the second combination, the fourth combination, the fifth combination, and the eighth combination are omitted.

  When the upper approximate set is created and stored in the predicate logical expression information storage area 312 (S77), the lower approximate set creation process is performed (S78, see FIG. 15). First, the relationship between the first condition attribute value set and the second condition attribute value set is determined for the specified combination of decision attributes (S101).

  In the example shown in FIGS. 26 and 27, in the third combination (“product purchase amount”, “current account”), the range of the first condition attribute value set is “X = 478155”, and the second condition attribute Since the range of the value set is “623 ≦ Y ≦ 300000”, there is no overlap. Therefore, it is determined as the first relationship, “1” is set in “Relation”, and nothing is set in “Overlap”. In the sixth combination (“communication transportation cost”, “current account”), the range of the first condition attribute value set is “623 ≦ X ≦ 6074”, and the range of the second condition attribute value set is “4200”. Since ≦ Y ≦ 478155 ”, the overlap is“ 4200 ≦ Z ≦ 6074 ”, which is determined as the second relationship. Therefore, “2” is set in “Relation”, “4200” is set in “Minimum Value” of “Overlap”, and “6074” is set in “Maximum Value”. In the seventh combination (“sales commission”, “current account”), the range of the first condition attribute value set is “150,000 ≦ X ≦ 200000”, and the range of the second condition attribute value set is “623 ≦ Since Y ≦ 478155 ”, since the first condition attribute value set is included in the second condition attribute value set, it is determined as the fourth relationship. Therefore, “4” is set in “Relation”, “150,000” is set in “Minimum value” of “Overlap”, and “200000” is set in “Maximum value”.

  When the relationship between the first condition attribute value set and the second condition attribute value set is determined for the specified combination of decision attributes (S101), a lower approximate set is created according to the relationship, and the condition attributes to be configured Are set to the “maximum value” and “minimum value” of the “lower approximation set” in the predicate formula information storage area 302 (S102 to S127).

  When the relationship is the first relationship (S102: YES), the first condition attribute value set is directly used as the lower approximation set, and therefore the “maximum value” and “minimum value” of the first cluster in the cluster information storage area 311 are used. Are set in the “maximum value” and “minimum value” of the “lower approximation set” in the predicate logical expression information storage area 312 (S105). Taking the third combination (“product purchase amount”, “current account”) as an example, the “maximum value” is “478155” and the minimum value is “478155”. Here, since the data having the condition attribute value “478155” belongs to the first cluster, it is set to “≦” instead of “<”, and the range of the lower approximation set is “478155 ≦ X ≦ 478155”. Become.

  Further, when the relationship is the second relationship (S102: NO, S103: YES), the set obtained by removing the overlap from the first condition attribute value set is set as the lower approximation set. A range obtained by excluding a range from “minimum value” to “maximum value” of “overlap” in the predicate formula information storage area 312 from a range from “minimum value” to “maximum value” of one cluster (S106). . Taking the sixth combination (“communication transportation cost”, “current account”) as an example, “4200 ≦ Z ≦ 6074” is deleted from “623 ≦ X ≦ 6074” and “minimum value” of the lower approximation set Is “623” and “maximum value” is “4200”. Here, since the data having the condition attribute value “623” belongs to the first cluster, it is set to “≦” instead of “<”, and the data having the condition attribute value “4200” belongs to the second cluster. Since the data is “<”, the range of the lower approximation set is “623 ≦ X <4200”. Although not particularly shown in FIG. 27, information indicating “<” or “≦” is stored in association with “maximum value” and “minimum value”.

  If the relationship is the third relationship (S102: NO, S103: NO, S104: YES), the condition attribute is a numeric type and not a character string type (S110: NO), so refer to FIG. As described above, the maximum value and the minimum value of the first cluster P are acquired (S111), and the value of the second condition attribute value included in the value (overlap) of the second cluster Q included in the first cluster P is acquired. (S112), these values are sorted (S113), and the number of values of the first cluster P included in the range between the smaller value and the larger value of the two adjacent values in the magnitude relationship. Is calculated (S114). Then, a range between two adjacent values having the largest number is determined (S115), and a set of elements of the first cluster P included in the range is set as a lower approximation set (S116).

  If the relationship is the fourth relationship (S102: NO, S103: NO, S104: NO), the first cluster P is included in the second cluster Q as described above with reference to FIG. In the case of (P⊆Q) (corresponding to “fourth relationship”), the maximum value and the minimum value of the first cluster P are acquired (S121), and the condition attribute is numeric and not a character string (S120). : NO), the value of the second condition attribute value included in the value (overlapping) of the second cluster Q included in the first cluster P is acquired (S122), these values are sorted (S123), and the magnitude relationship For the two adjacent values, the number of values of the first cluster P included in the range from the smaller value to the larger value is calculated (S124). Then, the range between the two adjacent values having the largest number is determined (S125), and the set of elements of the first cluster P included in the range is set as the lower approximate set (S126).

  Taking the seventh combination (“sales commission”, “current account”) in FIGS. 26 and 27 as an example, the maximum value of the first cluster is “200000”, and the minimum value is “150,000” (S121). The values of the second cluster included in the overlap are “150,000” and “160000” (S122). Therefore, since the respective relations are “150,000 = 150,000 <160000 <200000” (S123), if a range is specified for every two adjacent values, R1 (150,000 ≦ x ≦ 160000), R2 (160000 ≦ x ≦ 2 ranges) can be specified. The data of the first cluster having the condition attribute value within the range of R1 is one of “150,000”, and is one of “200000” within the range of R2 (S124). I want to select the range with the larger number of data, but since it is the same one, either can be selected. In this embodiment, the range with the larger numerical value is selected, and R2 (160000 ≦ x ≦ 200000) is selected (S125). Note that the larger numerical value may be selected, or the narrowest range may be selected. Here, since the data having the condition attribute value “200000” belongs to the first cluster, it is set to “≦” instead of “<”, and the data having the condition attribute value “160000” is the second cluster. Since the data belongs to “<”, the range of the lower approximation set is “160000 <x ≦ 200000” (S126).

  As described above, when the lower approximation set is created and stored in the lower approximation set column of the predicate logical expression information storage area 312 (S78), the validity evaluation value is calculated and the validity of the predicate logical expression information storage area 312 is calculated. It is stored in the sex evaluation value column (S79). This effectiveness evaluation value is calculated by the equation shown in Equation 6. Here, ρ = 2 and the third decimal place are regarded as effective digits and rounded off.

  When the combination of the specified determination conditions is the third combination (“product purchase amount”, “current account”), the number of event data having elements belonging to the upper approximate set is “1” of “478155” The number of event data having elements belonging to the lower approximate set is “1” of “478155”, the number of condition attributes constituting the combination of condition attributes is “1” of “current account withdrawal”, and all event data Since the number is “30”, the approximation accuracy is “1 ÷ 1 = 1” by dividing the number of event data having elements belonging to the lower approximation set by the number of event data having elements belonging to the upper approximation set. . Then, since the square root of the number of condition attributes “1” constituting the combination of condition attributes is “1”, the corrected approximation accuracy is “1 ÷ 1 = 1” by dividing the approximation accuracy “1”. Then, the effectiveness evaluation value is “1 × 1 ÷ 30 = 0.033” obtained by multiplying the approximate accuracy “1” by the number of event data having the key belonging to the lower approximate set divided by the total number of event data. Become.

  When the specified combination of determination conditions is the sixth combination (“communication transportation cost”, “current account”), the number of event data having elements belonging to the upper approximate set is “623”, “1320”, “4200”, “6074” “4”, the number of event data having elements belonging to the lower approximation set is “623”, “1320” “2”, the number of condition attributes constituting the combination of condition attributes is Since “current account withdrawal” is “1” and the total number of event data is “30”, the approximation accuracy is the number of event data having elements belonging to the upper approximation set and events having elements belonging to the lower approximation set. By dividing the number of data, “2 ÷ 4 = 0.5” is obtained. Then, since the square root of the number of condition attributes “1” constituting the combination of condition attributes is “1”, the corrected approximation accuracy is “0.5 ÷ 1 = 0 by dividing the approximation accuracy“ 0.5 ”. .5 ". Then, the effectiveness evaluation value is obtained by multiplying the approximate accuracy “0.5” by the number of event data having the key belonging to the lower approximate set divided by the total number of event data to obtain “0.5 × 2 ÷ 30 = 0. .033 ".

  When the specified combination of determination conditions is the seventh combination (“sales commission”, “current account”), the number of event data having elements belonging to the upper approximation set is “150,000”, “150,000”, “ “40000” of “100000” and “200000”, “1” of event data having elements belonging to the lower approximation set, “1” of “20000”, and the number of condition attributes constituting the combination of condition attributes is “current account withdrawal” "1" and the total number of event data is "30", the approximate accuracy is the number of event data having elements belonging to the upper approximation set and the number of event data having elements belonging to the lower approximation set. Thus, “1 ÷ 4 = 0.25”. The modified approximation accuracy is “0.25” /1=0.25 by dividing the approximation accuracy “1” because the square root of the number of condition attributes “1” constituting the combination of condition attributes is “1”. " Then, the effectiveness evaluation value is obtained by multiplying the approximation accuracy “1” by the number of event data having the key belonging to the lower approximation set, divided by the total number of event data, and “0.25 × 1 ÷ 30 = 0.003”. "

  When the validity evaluation value is calculated and stored in the validity evaluation value column of the cluster information storage area 312 (S79), a predicate logical expression is created and stored in the predicate logical expression column of the predicate logical expression information storage area 312 ( S80). The left side is the same as the case where the value of the condition attribute is a character string type, and is “journal (debit account name, credit account name)”, and the combination of determination attributes is set in parentheses. On the right side, “condition attribute 1 (A), relational expression 1 (A, value of condition attribute), relational expression 2 (A, value of condition attribute),... The relational expression is a character string indicating the relation between the variable A in () and the value of the condition attribute. In this embodiment, “A ≧ condition attribute value” is “greaterQual”, “A ≦ condition attribute value” is “lessEqual”, “A> condition attribute value” is “greaterThan”, “A <condition attribute value”. "Value" is expressed as "lessThan". A character string (Int / Float) indicating whether the value of the condition attribute is an integer type or a real number type is added to the end of the relational expression.

  In the example shown in FIG. 27, in the third combination of determination conditions, the determination attribute is (product purchase amount, current account), and the condition attribute is only “current account withdrawal”. Journal entries (product purchases, current account):-current account withdrawal (A), greaterEqualInt (A, 478155), lessEqualInt (A, 478155). " In the sixth combination of determination conditions, since the determination attribute is (communication transportation cost, current account) and the condition attribute is only “current account withdrawal”, the predicate logical expression is “journal (communication transportation cost, Current account):-current account withdrawal (A), greaterEqualInt (A, 623), lessEqualInt (A, 1320). " In the seventh combination of decision conditions, the decision attribute is (sales commission, current account) and the condition attribute is only “current account withdrawal”, so the predicate logical expression is “journal (sales commission, current account). ):-Current account withdrawal (A), greaterEqualInt (A, 160000), lessEqualInt (A, 200000).

  In this way, when predicate logical expressions are created for all combinations of decision attributes and stored in the predicate logical expression column of the cluster information storage area 312 (S77 to S80, S76: YES), the conditional attribute combination process ends. .

  Next, as an example in the case where the combination element of the condition attribute is a plurality of condition attributes, FIG. 28 to FIG. 32 are shown for the condition attribute combination processing in the case of a combination of “destination” and “information provision contract”. It explains using.

  FIG. 28 is a schematic diagram of the partial determination table 221 for the combination of the condition attributes “destination” and “information provision contract”, and FIG. 29 is a combination of the condition attributes “destination” and “information provision contract”. FIG. 30 is a schematic diagram of the partial decision table 223 for the combination of the condition attributes “destination” and “information provision contract”, and FIG. FIG. 32 is a schematic diagram of the cluster information storage area 321 of the RAM 101 that stores information on the first cluster and the second cluster for the combination of “partner” and “information provision contract”. It is a schematic diagram of a predicate logical formula information storage area 322 of the RAM 101 that stores set information and predicate logical formulas for a combination with an “information provision contract”.

  First, the partial determination table 221 (see FIG. 28) having only the condition attributes “partner” and “information provision contract” is created (S71). Then, as shown in FIG. 29, data (columns) having no values for the condition attributes “partner” and “information provision contract” are deleted from the partial determination table 221 (S72). In the example shown in FIG. 29, the data other than the 15th, 22nd, 25th, 26th, and 28th lines are deleted, and the partial determination table 222 is obtained.

  Next, the partial determination table 222 is sorted by the value of the determination attribute (S73), and the first cluster and the second cluster are determined (S74). The partial determination table 223 shown in FIG. 30 is a partial determination table in which the partial determination table 222 is sorted and the combinations of determination attributes are numbered.

  In the example shown in FIG. 30, in the first first cluster, the debit account item is “account expense” and the credit account item is “current account”. If the combination value of the condition attribute is indicated as (partner, information provision contract), the three data of (Mr. P, none), (Q University R professor, none), (Mr. S, none) belong. Yes. In the second first cluster, the debit account item is “sales commission” and the credit account item is “current account”, and the combination of condition attributes is (M research company, yes), (M research company, yes) ) Belong to two data.

  In S74, information about the generated first and second first clusters is stored in the cluster information storage area 321 of the RAM 101 (see FIG. 31). As shown in FIG. 31, since the data type of the condition attribute is “character string” for both “destination” and “information provision contract”, the “first condition attribute value set” is set for each condition attribute in the first cluster. In the second cluster, the “second condition attribute value set” is set.

  In the example shown in FIG. 31, in the first combination (accounting expense, current account), the data belonging to the first cluster is “Mr. P”, “Professor R at Q University”, “Mr. Is a first condition attribute value set. In the condition attribute “information provision contract”, a set having only “none” as an element is a first condition attribute value set. For data belonging to the second cluster, the condition attribute “destination” includes a set of only “M research company” as an element, and the condition attribute “information provision contract” has “Yes”. The set having only the element as the element is the second condition attribute value set.

  In the second combination (sales commission, current account), for the data belonging to the first cluster, the condition attribute “partner” includes a set having only “M survey company” as the first condition attribute value set. In the condition attribute “information provision contract”, a set having only “present” as an element is set as a first condition attribute value set. For the data belonging to the second cluster, the condition attribute “destination” is a set of three values of “Mr. P”, “Q Professor R” and “Mr. S” as the second condition attribute value set, In the condition attribute “information provision contract”, a set having only “none” as an element is a second condition attribute value set.

  As described above, when the first cluster and the second cluster are determined for each combination of determination attributes, and information related thereto is stored in the cluster information storage area 321 (S74), the predicate logic of the first determination attribute combination An expression is created and stored (S75 to S80), and then a combination predicate logical expression of the second decision attribute is created and stored (S75 to S80).

  First, when the combination of the first determination attributes is designated (S75), since all the combinations of the determination attributes have not been completed yet (S76: NO), the upper approximate set is determined for the combination of the first determination attributes. (S77). As shown in FIG. 32, in the first combination of decision attributes (interest expense, current account), the condition attribute “partner” has three values “P”, “Q University R Professor”, “M” S. “None” is set in the condition attribute “information provision contract”.

  Then, a lower approximate set creation process is performed (S78, see FIG. 15). The relationship between the first condition attribute value set and the second condition attribute value set is the first relationship because the condition attributes “destination” and “information provision contract” do not overlap and are independent relationships (S101). . And since it is the 1st relation (S102: YES), the 1st condition attribute value set is made into a lower approximation set (S105). As shown in FIG. 32, the value of the “lower approximation set” includes three values of “Mr. P”, “Professor R at Q University”, and “Mr. S” for each condition attribute. Is set, and in the condition attribute “information provision contract”, only “none” is set.

  Next, an effectiveness evaluation value is calculated for the first combination of decision attributes (S79). The number of event data having elements belonging to the upper approximation set is “3”, the number of event data having elements belonging to the lower approximation set is “3”, the number of condition attributes constituting the combination of condition attributes is “2”, and all events Since the number of data is “30”, the approximation accuracy is “3 ÷ 3 = 1” by dividing the number of event data having elements belonging to the upper approximation set by the number of event data having elements belonging to the lower approximation set. Become. Then, since the square root of the number of condition attributes “2” constituting the combination of condition attributes is “1.414”, the modified approximation accuracy is “1 ÷ 1.414 = 0 by dividing the approximation accuracy“ 1 ”. .707 ". Then, the effectiveness evaluation value is obtained by multiplying the approximate accuracy “0.707” by the number of event data having the key belonging to the lower approximate set divided by the total number of event data to obtain “0.707 × 3 ÷ 30 = 0. .071 ".

  Next, a predicate logical expression is created and stored in the predicate logical expression column of the predicate logical expression information storage area 322 (S80). The combination of the first determination attributes is (accounting expense, current account), the condition attributes are “partner” and “information provision contract”, and the lower approximation set is “P” for the condition attribute “partner”. “Mr.”, “Q University R Professor”, “Mr. S”, and the condition attribute “information provision contract” have “none” as an element. Therefore, the predicate logical expression is “Journaling (accounting expenses, current account):-Partner (A), member (A, [Mr. P, Professor R at Q University, Mr. S]), information provision contract (B), member ( B, [None]).

  Then, the process returns to S75, and the second decision attribute combination (sales fee, current account) has been made (S75), and since all the decision attribute combinations have not yet been completed (S76: NO), the designation is made. An upper approximate set is created for the combination of the determined second determination attributes (S77). In the example shown in FIG. 31 and FIG. 32, the condition attribute “partner” is set to “M survey company” and “information provision contract” is set to “present”.

  Next, a lower approximate set is created (S78, see FIG. 15). In the example shown in FIG. 31 and FIG. 32, the relationship between the first condition attribute value set and the second condition attribute value set is an independent relationship with no overlap between the condition attributes “destination” and “information provision contract”. Therefore, the first relationship is established (S101). And since it is the 1st relation (S102: YES), the 1st condition attribute value set is made into a lower approximation set (S105). As shown in FIG. 32, in the value of “lower approximation set”, “M survey company” is set for the condition attribute “partner” for each condition attribute, and “present” is set for the condition attribute “information provision contract”. Is set.

  Next, an effectiveness evaluation value is calculated for the first combination of decision attributes, and stored in the effectiveness evaluation value column of the predicate logical expression information storage area 322 (S79). The approximation accuracy is “2 ÷ 2 = 1” by dividing the number of event data having elements belonging to the upper approximation set by the number of event data having elements belonging to the lower approximation set. Then, since the square root of the number of condition attributes “2” constituting the combination of condition attributes is “1.414”, the modified approximation accuracy is “1 ÷ 1.414 = 0 by dividing the approximation accuracy“ 1 ”. .707 ". Then, the effectiveness evaluation value is obtained by multiplying the approximation accuracy “0.707” by the number of event data having key points belonging to the lower approximation set, divided by the total number of event data, to obtain “0.707 × 2 ÷ 30 = 0. .047 ".

  Next, a predicate logical expression is created and stored in the predicate logical expression column of the predicate logical expression information storage area 322 (S80). The second combination of decision attributes is (sales commission, current account), the condition attributes are “partner” and “information provision contract”, and the lower approximation set is “M” for the condition attribute “partner”. The “research company” and the condition attribute “information provision contract” have “present” as an element. Therefore, the predicate logical expression is “journal (sales fee, current account):-counterpart (A), member (A, [M research company]), information provision contract (B), member (B, [Yes])) . "

  Then, the process returns to S75, and the next decision attribute is specified (S75). However, since the process for all the decision attribute combinations has been completed (S76: YES), the conditional attribute combination process is terminated and the accounting is completed. Return to the journal entry rule creation process.

  As described above, the predicate logical expression when the combination of the condition attributes is “partner” and “information provision contract” is created.

  Next, journalizing processing 36 (see FIG. 11) for journalizing transaction data 35 using the predicate logical expression created in this way will be described with reference to the flowchart shown in FIG. FIG. 33 is a flowchart of the journalizing process.

  First, one transaction data is read from the transaction data stored in the transaction data storage area 125 of the hard disk device 120 of the computer 100 (S201). And transaction data can be read in S201, and if the processing of all transaction data is not completed (202: NO), one predicate logical expression stored in the predicate logical expression storage area 124 is read ( S203). Note that the predicate logical expressions are stored in descending order of effectiveness evaluation values, and are read out in descending order of effectiveness evaluation values. Then, the predicate logical expressions can be read, and if all the predicate logical expressions have not been processed (S204: NO), it is determined whether or not the transaction data satisfies the predicate logical expressions (S205).

  For example, the read transaction data is “Summary“ Today's Sales ”, Cash Deposit“ 123468 ”, No Value for Other Condition Attributes”, and the read predicate formula is “Journal (Sales Commission) , Checking account):-counterparty (A), member (A, [M research company]), information provision contract (B), member (B, [Yes]). In this case, since there is no value in the condition attribute “partner” or “information provision contract”, the value of the condition attribute “partner” is not “M survey company”, but the value of “information provision contract” is Since it is not “Yes”, this predicate logical expression is not satisfied.

  Therefore, since the predicate logical expression is not satisfied (S205: NO), the process returns to S203, and the next predicate logical expression is read (S203). Even in this case, if the predicate logical expression is not satisfied (S205: NO), the next predicate logical expression is read again (S203). Then, it is assumed that the predicate logical expression “Journal (cash, sales):-summary (A), member (A, [sales for today])” is read while the processing of S203 to S205 is repeatedly performed ( S203). This transaction data satisfies the predicate logical expression because the condition attribute “Summary” is “Today's sales” (S205: YES). Therefore, the transaction data, “cash” is stored in the “debit account item” as the determination attribute, and “sales” is stored in the “credit account item” in the journal result storage area 126 (S206). Then, the process returns to S201, and the next transaction data is processed.

  In this way, when the processing of S201 to S206 is repeated and the processing of all transaction data is completed (S202: YES), the journal entry processing is completed. Even if it is determined whether or not all the predicate logical expressions are satisfied (S205), if there is no predicate logical expression to satisfy (S204: YES), “debit account item” and “credit” Nothing is stored in “Account”.

  As described above, the journal result of the transaction data stored in the transaction data storage area 125 is stored in the journal result storage area 126.

  Therefore, as described above, journal rules (predicate formulas) that take into account the characteristics of the company can be created from the past journal results, so that journals according to the detailed rules and judgment materials existing in the company are performed. be able to. Therefore, even those who do not have knowledge about accounting can easily perform journalizing processing, which can contribute to cost reduction of accounting processing.

  In addition, since the common rules of those who have accounting knowledge can be reflected in the predicate formula creation, the predicate formula can be created more efficiently. In addition, detailed rules that exist in a company, rules with specific business partners, and specific judgment materials can be added as conditional attributes, so a predicate logical expression corresponding to the company that takes those specific rules into account is created. can do.

  The computer 100 corresponds to a “predicate formula creating device”, an “accounting journal rule creating device”, and an “accounting journal device”. The RAM 101 that stores the partial determination table corresponds to “event data storage unit”, the RAM 101 that stores the combination of condition attribute values corresponds to “all condition attribute combination storage unit”, and the predicate logical expression storage area 124 includes “ The journal result storage area 126 corresponds to “journal result storage means”.

  Then, the CPU 110 that performs the processing of S51 of the accounting journal rule creation processing of FIG. 13 corresponds to “event data input means”, and the CPU 110 that performs the processing of S57 corresponds to “all condition attribute combination creation means”, S58 to S60. The CPU 110 that performs the above process corresponds to “all condition attribute combination processing means”, and the CPU 110 that performs the processes of S63 and S64 corresponds to “predicate logical expression creation control means”. The CPU 110 that performs the process of the condition attribute combination process S74 in FIG. 14 corresponds to the “cluster creation unit”, the “first condition attribute value set creation unit”, and the “second condition attribute value set creation unit”. The CPU 110 that performs the processing corresponds to “relationship determination means”, the CPU 110 that performs the processing of S77 corresponds to “upper approximation set creation means”, the CPU 110 that performs the processing of S78 corresponds to “lower approximation set determination means”, and S79 The CPU 110 that performs the process of FIG. 8 corresponds to “approximation accuracy calculation means”, the CPU 110 that performs the process of S79 corresponds to “effective performance evaluation index calculation means”, and the CPU 110 that performs the process of S80 corresponds to “predicate logical expression creation means”. Equivalent to.

  The CPU 110 that performs the process of S105 of the lower approximate set creation process in FIG. 15 corresponds to the “first approximate set determining unit”, and the CPU 110 that performs the process of S106 corresponds to the “second approximate set determining unit”. The CPU 110 that performs the process of S117 corresponds to “third approximate set determining means”, the CPU 110 that performs the processes of S120 and S127 corresponds to “non-creating means”, and the CPU 110 that performs the process of S126 corresponds to “fourth approximate set determining means”. It corresponds to “means”. 33 corresponds to the “journaling means”, and the CPU 110 performing the process of S201 corresponds to “transaction data input means”.

  13 corresponds to the “event data input step”, S52 corresponds to the “event data storage step”, S57 corresponds to the “all condition attribute combination storage step”, and S57. Corresponds to “all condition attribute combination creation step”, S58 to S60 correspond to “all condition attribute combination processing step”, S63 and S64 correspond to “predicate logical expression creation control step”, and S64 corresponds to “predicate logic step”. Corresponds to “formula storage step”. 14 corresponds to the “cluster creation step”, corresponds to the “first condition attribute value set creation step”, corresponds to the “second condition attribute value set creation step”, and S76. This corresponds to the “relationship determination step”, S77 corresponds to the “upper approximation set creation step”, S78 corresponds to the “lower approximation set storage step”, S79 corresponds to the “approximation accuracy calculation step”, and S79 corresponds to “ This corresponds to the “effective performance evaluation index calculation step”, and S80 corresponds to the “predicate formula creation step”. 15 corresponds to the “first approximate set determining step”, S106 corresponds to the “second approximate set determining step”, and S110 to S117 are the “third approximate set determining step”. S120, S127 corresponds to a “non-creation step”, and S126 corresponds to a “fourth approximate set determination step”. 33 corresponds to the “transaction data input step”, S201 to S06 correspond to the “journal step”, and S206 corresponds to the “journal result storage step”.

  The predicate formula creation device, accounting journal rule creation device, accounting journal creation device, predicate formula creation program, accounting journal rule creation program, and accounting journal program of the present invention are limited to the above-described embodiments. However, it goes without saying that various changes can be made without departing from the scope of the present invention.

  In the above embodiment, transactions in accounting processing are taken as examples of events, rules for journalizing transaction data are created with a predicate logical expression, and transaction data is journalized using the predicate logical expression. Data handled in the predicate formula creation device and the predicate formula creation program is not limited to transaction data for accounting processing.

  In the above embodiment, the computer 100 functions as a “predicate formula creation device”, an “accounting journal rule creation device”, and an “accounting journalization device”, but the “accounting journal rule creation device” and the “accounting journal entry”. It goes without saying that the “device” may be a separate device. For example, the predicate formula created by the “accounting rule creation device”, that is, the “predicate formula creation device”, is another device via a storage medium such as a flexible disk or CD, or a network such as a LAN or the Internet. It may be stored in the “accounting journal device” and used. In addition, when performing journal processing in the “Accounting Journaling Program” of the “Accounting Journaling Device”, the predicate logical formula does not need to be stored in the “Accounting Journaling Device”, and other servers such as servers connected via the network A predicate logical expression stored in the apparatus may be read. In addition, a predicate formula creation program, an accounting journal rule creation program, an accounting journal program, and a predicate formula are stored in an ASP (Application Service Provider) device. Needless to say, the service may be connected and provided with a service.

  In the above embodiment, an upper approximation set is created for each combination of decision attributes (S77), a lower approximation set is created (S78), a validity evaluation value is calculated (S79), and a predicate logical expression is created. Then, after all the predicate logical expressions are created (S59: YES), the predicate logical expressions whose effectiveness evaluation values are lower than the reference value are deleted (S61 to S64). However, the process of creating and deleting the predicate logical expression is not limited to this flow, and may be after the validity evaluation value is created as long as the upper approximate set and the lower approximate set are created. For example, a predicate logical expression may be created before calculating the effectiveness evaluation value. In addition, the processing performed for each combination of decision attributes is only to create an upper approximation set and a lower approximation set, and after the upper approximation set and the lower approximation set of all decision attribute combinations are created, the information is read out. A predicate logical expression may be created. Further, when the effectiveness evaluation value is calculated, a comparison with the reference value is performed, and if it is lower than the reference value, the predicate logical expression may not be created.

  In the above embodiment, the journal result 37 is generated from the transaction data 35 by the journal process 36. However, the journal result is displayed on the display 114 after the journal process 36 is finished or the decision attribute is determined, You may make it confirm the journal entry result. When the operator determines that the journal result is valid, the instruction is input from the keyboard 115 or the mouse 108, and the journal result is stored in the journal result storage area 126 only when the input of the instruction is accepted. You may make it memorize | store. At that time, an instruction that the operator is not valid is input from the keyboard 115 or the mouse 108, and when the input is accepted, an input screen for inputting a decision attribute (debit account item, credit account item) is displayed. It may be displayed on the display 114 and the decision attribute input thereto may be stored in the journal result storage area 126. If the operator receives an input that is not valid, the operator continues to search for other predicate logical expressions that are satisfied by the transaction data, and displays decision attributes determined by the satisfied predicate logical expressions as candidates. You may do it. In this case, the attributes may be displayed in order from the decision attribute determined by the predicate logical expression having a high effectiveness evaluation value. In addition, even when there is no predicate logical expression that is satisfied, a screen for inputting a decision attribute is displayed on the display 114 after the journalizing process 36 is completed or after it is determined that it does not exist (S204: YES). The operator may be allowed to input a determination attribute. In addition, the journal result input by the operator may be stored separately so that it can be used as reference data for creating the company characteristic rule 33.

  The predicate logical expression creation device and the predicate logical expression creation program of the present invention are effective in classifying events having various attributes.

It is a schematic diagram which shows the concept of the data classification using predicate logic. It is a schematic diagram of the decision table 20 used by rough set theory. It is a schematic flowchart of rule creation. It is a schematic diagram of the partial determination table 21 for the condition attribute a of the determination table 20. It is an image figure of the cluster produced | generated from the partial determination table 21. FIG. It is an image figure of a lower approximation set when a condition attribute takes a character string. It is an image figure of a lower approximation set when a condition attribute takes a character string. It is an image figure of a lower approximation set when a condition attribute takes a numerical value. It is an image figure of a lower approximation set when a condition attribute takes a numerical value. It is an image figure of a lower approximation set when a condition attribute takes a numerical value. It is a related figure of data and processing. 2 is a block diagram showing an electrical configuration of a computer 100. FIG. It is a flowchart of an accounting journal rule creation process. It is a flowchart of the attribute combination process of an accounting journal rule creation process. It is a flowchart of lower approximation set creation processing of attribute combination processing. It is a schematic diagram of the determination table 501 created based on the past journal result 31 memorize | stored in the past journal result storage area 121. FIG. It is a schematic diagram of the determination table 502 after adding the common rule 32 and the company characteristic rule 33. It is a schematic diagram of the partial determination table 201 about the condition attribute "summary". It is a schematic diagram of the partial determination table 202 about the condition attribute "summary". It is a schematic diagram of the partial determination table 203 about the condition attribute "summary". It is a schematic diagram of the cluster information storage area 301 about the condition attribute “summary”. It is a schematic diagram of the predicate logical formula information storage area 302 for the condition attribute “summary”. It is a schematic diagram of the partial determination table 211 about the condition attribute "current account withdrawal". It is a schematic diagram of the partial determination table 212 for the condition attribute “current account withdrawal”. It is a schematic diagram of the partial determination table 213 about the condition attribute "current account withdrawal". It is a schematic diagram of the cluster information storage area 311 for the condition attribute “current account withdrawal”. It is a schematic diagram of the predicate logical expression information storage area 312 for the condition attribute “current account withdrawal”. It is a schematic diagram of the partial determination table 221 about the combination of condition attribute "destination" and "information provision contract". It is a schematic diagram of the partial determination table 222 about the combination of condition attribute "destination" and "information provision contract". It is a schematic diagram of the partial determination table 223 about the combination of condition attribute "destination" and "information provision contract". It is a schematic diagram of the cluster information storage area 321 for the combination of the condition attributes “partner” and “information provision contract”. FIG. 14 is a schematic diagram of a predicate logical expression information storage area 322 for a combination of condition attributes “partner” and “information provision contract”. It is a flowchart of a journal entry process.

Explanation of symbols

1 Event data 2 Event data 3 Event data 4 Event data 5 Event data 6 Event data 7 Event data 10 Event 12 Rule 13 Event data 20 Decision table 21 Partial decision table 30 Accounting journal rule creation process 31 Past journal result 32 Common rule 33 Company Characteristic rule 34 Predicate formula 35 Transaction data 36 Journal processing 37 Journal result 100 Computer 101 RAM
110 CPU
120 Hard disk device 121 Past journal result storage area 122 Common rule storage area 123 Enterprise characteristic rule storage area 124 Predicate logical expression storage area 125 Transaction data storage area 126 Journal result storage area 127 Program storage area 201 Partial determination table 202 Partial determination table 203 Part Decision table 211 Partial decision table 212 Partial decision table 213 Partial decision table 221 Partial decision table 222 Partial decision table 223 Partial decision table 301 Cluster information storage area 302 Predicate logical expression information storage area 311 Cluster information storage area 312 Predicate logical expression information storage area 321 Cluster information storage area 501 Decision table 502 Decision table P Cluster Q Cluster

Claims (14)

In a predicate logical expression creation device that creates a predicate logical expression that is a rule for classifying data for each event based on a predetermined attribute for an event having a plurality of attributes,
Event data input means for inputting event data, which is individual data for each event, having at least a determination attribute that is an attribute serving as a reference for classification and a condition attribute that is an attribute other than the determination attribute in a data item;
Event data storage means for storing a plurality of the event data input by the event data input means;
For the predetermined condition attribute of the event data stored in the event data storage means, the event data is a group of at least part of the event data having the same value of the determination attribute Cluster creation means for classifying the first cluster that is a set and the second cluster that is a set of at least part of the event data not included in the cluster;
When the value of the predetermined condition attribute of the event data included in the first cluster created by the cluster creation unit is a numerical value, all the events stored in the event data storage unit are stored. A set of elements having values not less than the minimum value and not more than the maximum value of the condition attribute value of the elephant data as a first condition attribute value set, and when the value of the predetermined condition attribute is a character string, A first condition attribute value set creating means that uses a set having condition attribute values as elements as a first condition attribute value set;
When the value of the predetermined condition attribute of the event data included in the second cluster created by the cluster creation means is a numerical value, all the events stored in the event data storage means A set of elements having values not less than the minimum value and not more than the maximum value of the condition attribute values of the image data as a second condition attribute value set, and when the value of the predetermined condition attribute is a character string, A second condition attribute value set creating means that uses a set having condition attribute values as elements as a second condition attribute value set;
A relationship determining means for comparing an element of the first condition attribute value set with an element of the second condition attribute value set and determining a relationship between the elements of the first condition attribute value set and the second condition attribute value set; ,
A predetermined subset of the first condition attribute value set is set as a lower approximate set based on a determination result by the relationship determining means, or a predicate logical expression is not generated without generating a lower approximate set. An approximate set determining means;
Lower approximation set storage means for storing the lower approximation set determined by the lower approximation set determination means;
For the lower approximation set stored in the lower approximation set storage means, the type of the predetermined condition attribute of the event data constituting the lower approximation set, the value of the predetermined condition attribute, and the value of the decision attribute A predicate logical expression creating means for creating a predicate logical expression indicating which value the decision attribute takes when the predetermined condition attribute takes a predetermined value;
A predicate logical expression creation device comprising: a predicate logical expression storage means for storing a predescription logical expression created by the predicate logical expression creation means. The relationship determination means compares the elements of the first condition attribute value set with the elements of the second condition attribute value set, and the relationship between the elements of the first condition attribute value set and the second condition attribute value set The first condition attribute value set and the second condition attribute value set do not overlap each other, the first condition attribute value set and the second condition attribute value set overlap each other and are included in each other A second relationship that is not included, a third relationship in which the second condition attribute value set is included in the first condition attribute value set, or a fourth relationship in which the first condition attribute value set is included in the second condition attribute value set. Determine if it ’s a relationship,
The approximate set determining means includes:
A first approximate set determining unit that uses the first condition attribute value set as a lower approximate set when the relationship determining unit determines that the first relationship is established;
A second approximate set determining unit that sets a set obtained by removing elements included in the overlap from the first condition attribute value set as a lower approximate set when the relationship determining unit determines that the second relationship is established;
When it is determined by the relationship determining means that the third relationship is established, and the value of the predetermined condition attribute is a character string, the second condition attribute value set is extracted from the first condition attribute value set. When the excluded set is a lower approximation set, and the value of the predetermined condition attribute is a numerical value, the maximum value and the minimum value of the first condition attribute value set and the first condition attribute value set are included. Of the first condition attribute value set included in a range indicated by a smaller value and a larger value of two adjacent values of all magnitude relations A third approximation set determining unit that calculates the number of elements and sets a set of numbers included in a range specified by two adjacent values having the largest number as a lower approximation set;
When it is determined by the relationship determining means that the relationship is the fourth relationship, if the value of the predetermined condition attribute is a numerical value, the maximum value and the minimum value of the first condition attribute value set and the overlap Among the values of the second condition attribute value set included in the first condition attribute value included in a range indicated by a smaller value and a larger value of two adjacent values of all magnitude relationships A fourth approximate set determining unit that calculates the number of elements of the set and sets a set of numbers included in a range specified by two adjacent values having the largest number as a lower approximate set;
A non-creating unit that does not create a predicate logical expression when the value of the predetermined condition attribute is a character string when the relationship determining unit determines that the fourth relationship is the fourth relationship; The predicate logical expression creation device according to claim 1. There are a plurality of the condition attributes,
All condition attribute combination creating means for creating all combinations of the condition attributes of the event data stored in the event data storage means;
All condition attribute combination storage means for storing the combination created by the all condition attribute combination creation means;
For all combinations stored in the all condition attribute combination storage means, the cluster creation means, the first condition attribute value set creation means, the second condition attribute value set creation means, the relationship determination means, 3. The predicate logical expression creation device according to claim 1, further comprising: all condition attribute combination processing means for executing the lower approximate set determination means and the previous description word logical expression creation means. An upper approximation set creating means having the first condition attribute value set as an upper approximation set;
The upper approximate event data number that is the number of the event data in which the value of the predetermined condition attribute is included in the upper approximate set stored in the approximate set storage means, and the predetermined condition in the lower approximate set Approximate accuracy calculation that calculates the number of lower approximate event data, which is the number of event data including attribute values, and calculates the approximate accuracy that is the value obtained by dividing the number of lower approximate event data by the number of upper approximate event data Means,
For each combination, based on the number of lower approximate event data of the combination, the approximate accuracy calculated by the approximate accuracy calculation means, and the number of event data stored in the event data storage means An effective performance evaluation index calculating means for calculating an effective performance evaluation index,
A predicate that prevents the pre-description logical formula creation means from creating a pre-description logical formula for the combination in which the value of the effective performance evaluation index calculated by the effective performance evaluation calculation means is less than or equal to a predetermined value The predicate logical expression creation device according to claim 1, further comprising: logical expression creation control means.   In the event data stored in the event data storage means, when the value of the predetermined condition attribute is not stored, the cluster creation means, the first condition attribute value set creation means, the first 5. The predicate logical expression creation according to claim 1, wherein the two-condition attribute value set creation means, the relation determination means, the lower approximation set determination means, and the previous description word logical expression creation means are not operated. apparatus. The predicate logical expression creation device according to any one of claims 1 to 5,
The event is a transaction in accounting;
The event data is transaction data relating to individual transactions that occur in the accounting process,
The types of decision attributes are debits and credits, the elements of which are account items,
The condition attribute is various related items related to the accounting transaction, and its elements are contents related to the related items,
An accounting journal rule creation device for creating a predicate logical formula for determining which account item the debit and credit as the determination attributes are when a predetermined condition attribute takes a predetermined value. Transaction data input means for inputting transaction data relating to individual transactions that occur in accounting processing;
The debit of the transaction data input by the transaction data input means is the determination attribute according to the previous description word logical expression created by the previous description word logical expression creation means of the accounting journal rule creation device according to claim 6. And journalizing means to determine the credit account,
An accounting journalizing apparatus comprising: a journal result storage unit that stores the debit and credit account items, which are the determined attributes determined by the journalizing unit, in association with the transaction data. A predicate formula creation program for causing a computer to execute an operation for creating a predicate formula, which is a rule for classifying data for each event based on a predetermined attribute, for an event having a plurality of attributes,
An event data input step for inputting event data, which is individual data for each event, having at least a determination attribute that is an attribute serving as a classification criterion and a condition attribute that is an attribute other than the determination attribute in a data item;
An event data storage step of storing a plurality of the event data input in the event data input step in an event data storage means;
For the predetermined condition attribute of the event data stored in the event data storage means, the event data is a collection of at least part of the event data having the same value of the determination attribute A cluster creation step of classifying the first cluster that is a set of the first cluster and the second cluster that is a set of at least part of the event data not included in the cluster;
If the value of the predetermined condition attribute of the event data included in the first cluster created by the cluster creation step is a numerical value, all the events stored in the event data storage means are stored. A set of elements having values not less than the minimum value and not more than the maximum value of the condition attribute value of the elephant data as a first condition attribute value set, and when the value of the predetermined condition attribute is a character string, A first condition attribute value set creation step in which a set having condition attribute values as elements is a first condition attribute value set;
When the value of the predetermined condition attribute of the event data included in the second cluster created by the cluster creation step is a numerical value, all the events stored in the event data storage means are stored. A set of elements having values not less than the minimum value and not more than the maximum value of the condition attribute values of the image data as a second condition attribute value set, and when the value of the predetermined condition attribute is a character string, A second condition attribute value set creation step in which a set having condition attribute values as elements is a second condition attribute value set;
A relationship determining step of comparing an element of the first condition attribute value set with an element of the second condition attribute value set and determining a relationship between the elements of the first condition attribute value set and the second condition attribute value set; ,
Based on the determination result in the relationship determination step, a predetermined subset of the first condition attribute value set is set as a lower approximate set, or a predicate logical expression is not generated without generating a lower approximate set. An approximate set determination step;
A lower approximation set storage step for storing the lower approximation set determined in the lower approximation set determination step in a lower approximation set storage means;
For the lower approximation set stored in the lower approximation set storage means, the type of the predetermined condition attribute of the event data constituting the lower approximation set, the value of the predetermined condition attribute, and the value of the decision attribute A predicate logical expression creating step for creating a predicate logical expression indicating which value the decision attribute takes when the predetermined condition attribute takes a predetermined value;
A predicate logical expression creation program for causing a computer to execute a predicate logical expression storage step of storing in a predicate logical expression storage means the predicate logical expression created by the predicate logical expression creation step. The relationship determining step causes the computer to compare an element of the first condition attribute value set with an element of the second condition attribute value set, and to determine an element between the first condition attribute value set and the second condition attribute value set. The first condition attribute value set and the second condition attribute value set have no overlap, and the first condition attribute value set and the second condition attribute value set have an overlap. Second relation not included in each other, third relation in which the second condition attribute value set is included in the first condition attribute value set, or the first condition attribute value set is included in the second condition attribute value set To determine if it is any of the fourth relationships,
The approximate set determination step includes:
A first approximate set determining step in which the first condition attribute value set is a lower approximate set when it is determined in the relationship determining step that the first relationship is;
A second approximate set determining step in which a set obtained by removing elements included in the overlap from the first condition attribute value set as a lower approximate set when it is determined that the second relationship is determined in the relationship determining step;
If it is determined in the relationship determination step that the third relationship is established, and the value of the predetermined condition attribute is a character string, the second condition attribute value set is extracted from the first condition attribute value set. When the excluded set is a lower approximation set, and the value of the predetermined condition attribute is a numerical value, the maximum value and the minimum value of the first condition attribute value set and the first condition attribute value set are included. Of the first condition attribute value set included in a range indicated by a smaller value and a larger value of two adjacent values of all magnitude relations Calculating a number of elements, and a third approximation set determining step in which a set of numbers included in a range specified by two adjacent values having the largest number is set as a lower approximation set;
If it is determined in the relationship determination step that the relationship is the fourth relationship, and the value of the predetermined condition attribute is a numerical value, the maximum value and the minimum value of the first condition attribute value set and the overlap Among the values of the second condition attribute value set included in the first condition attribute value included in a range indicated by a smaller value and a larger value of two adjacent values of all magnitude relationships Calculating the number of elements of the set, and a fourth approximate set determining step in which the set of numbers included in the range specified by two adjacent values having the largest number is the lower approximate set;
When it is determined that the relationship is the fourth relationship in the relationship determination step, if the value of the predetermined condition attribute is a character string, a non-creation step that does not create a predicate logical expression is 9. The predicate logical expression creation program according to claim 8, wherein the predicate logical expression creation program is executed. There are a plurality of the condition attributes,
An all condition attribute combination creating step for creating all combinations of the condition attributes of the event data stored in the event data storage means;
An all-condition attribute combination storage step for storing the combination created by the all-condition attribute combination creating step in the all-condition attribute combination storage means;
For all combinations stored in the all condition attribute combination storage unit, the cluster creation step, the first condition attribute value set creation step, the second condition attribute value set creation step, the relationship determination step, 10. The predicate formula creation program according to claim 8 or 9, causing a computer to execute the lower approximate set determination step and the precondition word formula creation step. An upper approximation set creating step in which the first condition attribute value set is an upper approximation set;
The upper approximate event data number that is the number of the event data in which the value of the predetermined condition attribute is included in the upper approximate set stored in the approximate set storage means, and the predetermined condition in the lower approximate set Approximate accuracy calculation that calculates the number of lower approximate event data, which is the number of event data including attribute values, and calculates the approximate accuracy that is the value obtained by dividing the number of lower approximate event data by the number of upper approximate event data Steps,
For each combination, based on the number of lower approximate event data of the combination, the approximate accuracy calculated by the approximate accuracy calculation step, and the number of event data stored in the event data storage means An effective performance evaluation index calculating step for calculating an effective performance evaluation index,
A predicate that does not create a pre-description logical formula in the pre-description logical formula creation step for the combination in which the value of the effective performance evaluation index calculated by the effective performance evaluation calculation step is less than or equal to a predetermined value The predicate logical expression creation program according to any one of claims 8 to 10, wherein the computer executes the logical expression creation control step.   In the event data stored in the event data storage means, when the value of the predetermined condition attribute is not stored, the cluster creation step, the first condition attribute value set creation step, the first The predicate logic according to any one of claims 8 to 11, wherein the computer does not execute a two-condition attribute value set creation step, the relation determination step, the lower approximation set determination step, and the previous description word logical expression creation step. Formula creation program. In the predicate logical expression creation program according to any one of claims 8 to 12,
The event is a transaction in accounting;
The event data is transaction data relating to individual transactions that occur in the accounting process,
The types of decision attributes are debits and credits, the elements of which are account items,
The condition attribute is various related items related to the accounting transaction, and its elements are contents related to the related items,
Accounting journal rule characterized in that when a predetermined condition attribute takes a predetermined value, the computer executes creation of a predicate logical expression that determines which account item the debit and credit as the determination attributes are Creation program. A transaction data entry step for entering transaction data for individual transactions that occur in accounting;
The debit which is the decision attribute is the transaction data input in the transaction data input step according to the predescription word formula created by the predescription word formula creation step of the accounting journal entry rule creation program according to claim 13. Journal steps to determine the credit and credit accounts;
An accounting journal program that causes a computer to execute a journal result storage step of storing the debit and credit account items that are the determined attributes determined in the journal step in association with the transaction data.

Images