KR20240032759A - Injury judgement system - Google Patents

Abstract

An objective of the present invention is to provide an injury disease determination system capable of automatically determining an injury disease to be preceded when examining personal insurance subscription, minimizing errors, and ensuring reliability. The present invention comprises: an accident history generation module for collecting accident history information of an insured person described in notification history information received online to generate accident history information for each injury disease; an accident history classification module for classifying a transmission path of the accident history information for each injury disease according to a plurality of classification criteria; an injury disease notification probability granting module for granting an injury and illness notification probability of each injury disease code; an injury disease notification probability prediction module for predicting an injury disease notification probability for each injury disease code; an injury disease selection module for selecting an injury disease code to be notified of an injury disease; and an injury disease final determination module for determining and outputting a final injury disease with respect to the injury disease code and the injury disease notification probability. According to an embodiment of the present invention, an injury disease determination system predicts an injury disease notification probability for all injury disease codes included in accident history information and analyzes the correlation of each predicted injury disease code to select an injury disease, thereby minimizing an error in automatic judgment of an injury disease compared to an existing artificial intelligence network which simply determines an injury disease based on the notified injury disease.

Description

INJURY JUDGEMENT SYSTEM}

The present invention relates to a personal insurance subscription screening system, and in particular, to an injury/disease judgment system that automatically determines injuries and diseases that must be preceded when screening personal insurance subscription.

In general, in the case of long-term insurance, depending on the subject matter that can be selected, there are life insurance, water insurance, and insurance products that combine water and people.

Personal insurance (人保險) is insurance in which the insurer agrees to pay a certain amount of money and other benefits to the insured in the event of an accident that threatens the life or body of the insured, and the policyholder agrees to pay compensation for this. Insurance can be said to be insurance that promises to pay compensation for accidents that occur with respect to goods or other property.

In order to subscribe to insurance, the subject matter of the insurance must first be reviewed. In order to shorten business hours and increase customer convenience, many insurance companies are introducing AI (artificial intelligence) service systems to each business.

For example, the 'AI contract system' introduced by company A is a type of screening system that increases the number of cases that can be subscribed only through computer screening by training AI to immediately approve contract types without additional confirmation, and the 'insurance consulting service' introduced by company B is a consultant It is a consultation system that automates the simple work processing of reviewers, and Company C's 'Insurance AI Automatic Review System' is a system that applies AI to insurance payment work.

In this way, AI systems are being introduced to improve work efficiency, reduce costs, and increase customer convenience by applying AI technology to each insurance business, but the biggest problem is whether the reliability of the introduced AI system can be secured.

For example, when AI technology is applied to personal insurance screening, the first thing to be determined is the injury or disease, and then the collateral judgment must be made to determine whether to accept, add a premium, or reject the insurance application. In this series of judgment procedures, determining whether a person is injured or sick must come first, but judging whether a person is injured or sick by simply applying rule-based or AI technology is not that easy. The reason is that the number of illness codes registered in the Korean Standard Classification of Diseases/Causes (KCD7) is very large, and there are several illness codes in prescriptions issued for specific diseases (e.g., for colds, upper respiratory tract infection code, gastrointestinal disorder code, Mental disorder codes may be included), so it is impossible to automatically and mechanically determine the illness for screening by simply recognizing the notified illness code and the illness code included in the prescription. Accordingly, there is a need to develop a reliable injury and disease determination system that automatically determines injuries and diseases that must take precedence when reviewing personal insurance subscription.

Republic of Korea Patent Publication No. 10-2020-0138085 Republic of Korea Patent Publication No. 10-2019-0105211 Republic of Korea Patent Publication No. 10-1955361

Accordingly, the present invention is an invention created in accordance with the above-mentioned necessity, and the main purpose of the present invention is to provide an injury and disease determination system that automatically determines injuries and diseases that must be preceded when examining personal insurance subscription, while minimizing errors and ensuring reliability. ,

Furthermore, we provide an artificial intelligence-based injury judgment system that can automatically determine injury or disease quickly, accurately, and reliably, regardless of the personal insurance holder's work skills and experience.

The injury determination system according to an embodiment of the present invention for achieving the above-described purpose is an injury determination system executable on a computer system,

a thinking ability generation module that generates thinking ability information for each injury by collecting the thinking ability information related to the injury and disease described in the notification history information of the insurance subscriber submitted online;

A thinking ability classification module that classifies the transmission path of the thinking ability information for each disease according to a plurality of classification criteria;

an injury notification probability granting module that grants an injury notification probability for each injury code by applying the thinking ability information for each injury transmitted according to the classification criteria to a predefined rule;

An injury notification probability prediction module that predicts the injury notification probability for each injury code by applying the thinking ability information for each injury delivered according to the classification criteria to a logistic regression model;

an injury selection module that selects an injury code to be notified based on a correlation analysis result of each injury code for which the injury notification probability is predicted;

It is characterized in that it includes; a final injury judgment module that determines and outputs a final injury with respect to the injury code and injury notification probability transmitted from the injury disease notification probability granting module and the injury disease selection module.

Furthermore, in the above-described injury judgment system, the injury final judgment module is,

Another feature is that if the injury notification probability of the injury code transmitted from the injury selection module exceeds the specified value matching the preset loss ratio, it is judged as the final injury and output,

The regulation value matched to the above loss ratio is characterized as an injury disease notification probability value determined by considering the first year loss rate for each injury and disease.

Meanwhile, in the injury judgment system of the above-described configuration, the thinking ability classification module is,

It is characterized in that it includes first classification criteria for classifying severe diseases and mild diseases among the thinking ability information for each disease and transmitting it to the disease notification probability granting module,

Furthermore, it further includes second classification criteria that are different from the first classification criteria, but in this case, the injury disease notification probability prediction model includes a logistic regression model matching each of the second classification criteria to predict the injury disease notification probability. Another characteristic is that

According to the above-mentioned technical problem solving means, the injury judgment system according to an embodiment of the present invention collects injury-related thinking ability information related to the insured person and generates thinking ability information for each disease, so information about injury codes included in prescriptions Not only can it be included in thinking ability information, but the injury notification probability is predicted for all injury codes included in thinking ability information, and injury selection is performed by analyzing the correlation between each predicted injury code. Compared to existing artificial intelligence networks that make judgments, it has the effect of minimizing errors in automatic judgment of injuries and illnesses.

Furthermore, the injury and disease determination system according to an embodiment of the present invention automatically determines the injury or disease that must be preceded when screening for personal insurance subscription, so not only can it reduce the reviewer's workload, but can also quickly process the personal insurance subscription screening process. There is an advantage.

Figure 1 is an example of the surrounding configuration of an injury judgment system 200 using artificial intelligence according to an embodiment of the present invention.
Figure 2 is an example configuration of an injury judgment system 200 using artificial intelligence according to an embodiment of the present invention.
Figure 3 is an exemplary flow diagram of an injury or disease determination process according to an embodiment of the present invention.
Figure 4 is an example of thinking ability information for each injury or disease according to an embodiment of the present invention.
Figure 5 is an example of classification criteria according to an embodiment of the present invention.
Figure 6 is an example of learning data and output data of the injury notification probability prediction module 240 according to an embodiment of the present invention.
Figure 7 is an example graph of loss ratio and injury notification probability for each injury and disease according to an embodiment of the present invention.

Specific structural or functional descriptions of the embodiments according to the concept of the present invention disclosed in this specification are merely illustrative for the purpose of explaining the embodiments according to the concept of the present invention, and the embodiments according to the concept of the present invention are It may be implemented in various forms and is not limited to the embodiments described herein.

In addition, since the embodiments according to the concept of the present invention can make various changes and have various forms, the embodiments will be illustrated in the drawings and described in detail in this specification. However, this is not intended to limit the embodiments according to the concept of the present invention to specific disclosed forms, and includes all changes, equivalents, or substitutes included in the spirit and technical scope of the present invention. In addition, when describing embodiments of the present invention, if it is determined that specific descriptions such as related known functions or configurations may unnecessarily obscure the gist of the present invention, detailed descriptions thereof will be omitted.

In addition, the injury determination system using artificial intelligence according to an embodiment of the present invention may be a set of codes constituting an application program or a control program of a computer system, and may include a predetermined code and hardware resources for executing the predetermined code. It may be implemented in the form of a system composed of multiple modules, which represent logical units.

Hereinafter, in order to enable those skilled in the art to easily practice the present invention, preferred embodiments of the present invention will be described in detail with reference to the attached drawings.

Figure 1 illustrates the peripheral configuration of an injury judgment system using artificial intelligence according to an embodiment of the present invention.

As shown, the injury determination system 200 according to an embodiment of the present invention is a system that can be installed and executed on a server computer system operated by a financial institution, and is capable of transmitting and receiving data to and from a plurality of planner terminals 100 through a communication network. Of course, data can also be transmitted and received with the insurance claim details DB (300), which contains thinking ability information. It is desirable to understand this insurance claim details DB 300 as a server that stores, manages, and shares details of insurance claims made to various financial institutions, for example, various insurance companies.

The injury judgment system 200 recognizes the related injury from the notification history information (preferably interpreted as including accident history) written or attached to the insurance application form received through the planner terminal 100, and recognizes the By collecting the information on the insured's thinking skills recorded in the injury and disease notification history information, predicting the probability of notification of each injury and analyzing the correlation, it automatically determines the final injury to be notified.

The configuration of this injury/disease determination system 200 will be described in more detail with reference to FIG. 2. Each module or some modules of the injury/disease determination system 200, which will be described below, may be implemented as modules using artificial intelligence.

Figure 2 illustrates the configuration of an injury judgment system 200 using artificial intelligence according to an embodiment of the present invention, Figure 4 shows thinking ability information for each injury and disease according to an embodiment of the present invention, and Figure 5 shows the present invention. Classification criteria according to an embodiment of the present invention, Figure 6 shows the learning data and output data of the injury notification probability prediction module 240 according to an embodiment of the present invention, and Figure 7 shows the loss rate and injury rate for each injury and disease according to an embodiment of the present invention. These are examples of notification probability graphs.

First, as shown in FIG. 2, the injury determination system 200 according to an embodiment of the present invention,

A thinking ability generation module (210) collects the thinking ability information of the insured as well as the thinking ability information related to injury and disease described in the online application or entered notification history information from the insurance claim details DB (300) and generates thinking ability information for each injury and disease as shown in FIG. 4. )class,

A thinking ability classification module 220 that classifies the transmission path of the thinking ability information for each injury or disease according to a plurality of classification criteria as illustrated in FIG. 5,

An injury notification probability granting module 230 that applies the thinking ability information for each illness delivered according to the above classification criteria to a predefined rule to grant an illness notification probability of an illness code corresponding to a serious illness or a very mild illness;

An injury notification probability prediction module 240 that predicts the injury notification probability for each injury code by applying the thinking ability information for each injury delivered according to the above classification criteria to a logistic regression model;

An injury selection module 250 that selects an injury code to be notified based on the correlation analysis results of each injury code for which the injury notification probability is predicted;

It includes an injury final determination module 260 that determines and outputs a final injury based on the injury code and injury notification probability transmitted from the injury notification probability granting module 230 and the injury disease selection module 250.

Among the above-described configurations, the thinking ability generation module 210 recognizes the natural language described in the notification history information and collects the thinking ability information related to the recognized insured person's injury from the insurance claim details DB 300, but does not overlap within a certain period for the same injury code. Thinking ability information for each injury, as shown in FIG. 4, can be generated by removing thinking ability information and adding thinking ability information for the same injury code that relapsed within a predetermined period of time, and can be used to determine illness.

The thinking ability classification module 220 includes first classification criteria for classifying severe injuries and very mild injuries among the thinking ability information for each illness generated by the thinking ability generation module 210 and transmitting it to the injury notification probability granting module 230.

In Figure 5, the first classification criteria are 'unable to judge' standard, 'notice of dental collateral', 'notice of fetal insurance' standard, 'notice of outpatient visit for m days within n years but in case of severe condition' standard, and 'notice of outpatient visit for m days within n years' standard. It is a non-severe case and includes the standard of having a notification rate of 'p% or more'. In these standards, the period (years and days) and notification rate figures (p, p are natural numbers, n and m are natural numbers) can be varied according to in-house standards.

It is desirable to understand the notification rate as the ratio of notified cases among the total data analyzed (learned) to create classification standards. In addition, among the terms used in the present invention, the 'injury notification probability' can be defined as the probability of notifying the examiner of an injury and making a judgment on the injury, and if this injury notification probability exceeds a predetermined regulation value, the final injury notification probability can be defined. It can be judged and displayed and output.

Furthermore, the thinking ability classification module 220 further includes second classification criteria that are different from the first classification criteria. As shown in Figure 5, the second classification criteria are 'notification upon surgery or hospitalization' standard, 'outpatient notification on m days within n years with notification rate below p%', and 'notification upon surgery, long-term hospitalization or recurrence'. It may be the standard, 'ultra-long hospitalization or frequent recurrence notification' standard, or 'no notification' standard. (n, m, p are natural numbers)

The injury notification probability prediction model 240 includes a logistic regression model matching each of the second classification criteria, so that the injury notification probability can be predicted for the injury code classified according to each second classification standard. For reference, the logistic regression model learns the thinking ability information for each injury (which can be considered a type of AI learning data) as shown in FIG. 6 and outputs an injury notification probability value for each injury (indicated as an analysis item value in FIG. 6). It can be said to be a pre-designed model. In some cases, the logistic regression model includes four types of thinking ability information for each injury as shown in FIG. 6, which have a great influence on the probability of notification of injury: number of days hospitalized, number of days outpatient, elapsed time from accident to present, and whether or not surgery is performed. You can also use a pre-designed model to pre-learn information and output an injury notification probability value.

On the other hand, among the above-described configurations, the injury final judgment module 260 determines and outputs the final injury if the injury code transmitted from the injury notification probability granting module 230 is a severe injury (100% injury notification probability). In case of an injury (0% probability of notification of injury), the notification of injury may be ignored.

As another embodiment that can be modified, the injury final judgment module 260 determines the injury as the final injury when the injury notification probability of the injury code transmitted from the injury selection module 250 exceeds the specified value matching the preset loss ratio. You can also print it out.

In this case, the regulation value matching the above loss rate may be an injury disease notification probability value determined by considering the first year loss rate for each injury and disease. The above first-year loss ratio and injury notification probability values are values that can vary depending on the insurance company's internal regulations.

Hereinafter, the operation of the injury determination system 200 including the above-described configuration will be described in detail with reference to FIGS. 2 to 7.

Figure 3 illustrates an injury/disease determination processing flowchart according to an embodiment of the present invention.

Referring to FIG. 3, first, the thinking ability generation module 210 constituting the injury/disease determination system 200 receives an insurance application form online through the planner terminal 100 (step S10).

The insurance application form submitted online contains notification history information, including the insured person's (customer) previous accident history. Accordingly, the thinking ability generation module 210 recognizes the natural language described in the notification history information and collects the recognized injury and illness and the insured person's injury-related thinking ability information from the insurance claim history DB 300, but the thinking ability information that overlaps within a certain period for the same injury code is is removed and the thinking ability information of the same recurrent injury code is added to generate thinking ability information for each injury as shown in FIG. 4 (step S20).

As shown in Figure 4, the thinking ability information for each disease includes the number of days hospitalized due to the disease (code), number of outpatient days, whether surgery was performed, number of surgeries, recurrence, elapsed time, insurance money (payment amount), gender, age, and last treatment. Information such as date, onset date, and accident factors (disease, injury, traffic accident) is included. This information on thinking ability by illness may include, in addition to the notified illness, illness codes based on prescriptions issued in relation to the notified illness.

When thinking ability information for each injury is generated for the customer by the thinking ability generation module 210, the thinking ability classification module 220 checks whether the generated thinking ability information for each injury corresponds to the classification criteria shown in FIG. 5 to determine the thinking ability for each injury. Classifies the information transmission path (step S30).

These classification standards include a first classification standard for extracting serious illnesses and very mild illnesses, and a second classification standard for extracting illnesses that are not serious illnesses and very mild illnesses. In the embodiment of the present invention, these classification criteria include 14 classification criteria as shown in FIG. 5, but the number of classification criteria may be variable depending on the amount of training data secured and the level of reliability.

For reference, the first classification criteria for extracting severe and mild morbidity are as shown in Figure 5: 'unable to judge' standard, 'notification of dental collateral' standard, 'notification of fetal insurance' standard, 'notification of outpatient visit for m days within n years, but severe disease. The 'case' standard includes the standard of 'notification of outpatient care on m days within n years, but in non-severe cases, and the notification rate is more than p%'.

If the thinking ability information for each illness transmitted from the thinking ability generation module 210 corresponds to the first classification standard for classifying a serious illness or a very mild illness in the thinking ability classification module 220, the thinking ability information for each illness is provided to the injury notification probability granting module ( 230).

The injury notification probability granting module 230 applies the delivered thinking ability information for each injury to a predefined rule to grant an injury notification probability for each injury code (step S40). For example, if the thinking ability information satisfies the rule for a severe illness, the probability of notifying an illness is given as 100%, and if the information for a very mild illness satisfies the rule for a very mild illness, the probability of notifying an illness is given as 0%. The probability of notifying a sick person is given to a sick person. For reference, a 100% injury/disease notification probability can be interpreted to mean rejection of an application or review, and a 0% injury/disease notification probability can be interpreted to mean the application is approved as standard without being processed by the reviewer.

On the other hand, if the thinking ability information for each injury transmitted from the thinking ability generation module 210 does not correspond to the first classification criteria, that is, if it corresponds to any one of the second classification criteria, the thinking ability information for each injury is transmitted to the injury notification probability prediction module 240. is passed on. Since the types of the second classification criteria were described above, they are omitted below.

The injury notification probability prediction module 240 applies the delivered thinking ability information for each injury to a logistic regression model to predict the injury notification probability for each injury code (step S50). This injury disease notification probability prediction model 240 includes a logistic regression model matching each of the second classification criteria and predicts the injury disease notification probability.

More specifically, the logistic regression model matching each of the second classification criteria includes the number of hospitalization days, number of outpatient days, elapsed time from accident to present, and When four types of information on whether or not to have surgery are input, the injury notification probability for the corresponding injury code is predicted and output. Of course, it is also possible to predict the injury notification probability for a specific injury code using a logistic regression model learned by setting the cause value and result value as illustrated in FIG. 6 as AI learning data. In Figure 6, the analysis item value '100' means a 100% probability of notification of injury or disease, meaning rejection of the subscription, and the analysis item value 'null' means a 0% probability of notification of injury or disease, which can be interpreted to mean approval as standard.

If the injury notification probability for the injury codes is predicted as above, then the injury selection module 250 notifies the injury based on the correlation analysis results of each injury code for which the injury notification probability is predicted (e.g., injury A, injury B). Select the injury code to be used (step S60). In other words, the injury selection module 250 selects the injury to be notified to the reviewer based on the Association Rule (Confidence, Support and injury notification probability statistics) of the injury pair. For reference, Support indicates that injuries A and B occurred simultaneously among all injuries, and Confidence indicates the frequency of occurrence of both A and B among the frequency of occurrence of A or B.

For example, when the following diseases occur simultaneously, the disease screening module 250 sets the disease code to be notified to urinary stones when the disease notification probability is found to be 52.7% and 23.2% as a result of the correlation analysis for urinary stones and hydronephrosis, respectively. If the disease notification probability is found to be 89.4% and 70.1% as a result of the correlation analysis for glaucoma and macular degeneration, the disease code to be notified can be selected as glaucoma.

As described above, if the injury notification probability is assigned to a serious illness or a very mild illness, or if the illness notification probability is predicted for an illness other than a severe illness or a very mild illness, the injury final judgment module 260 grants the injury notification probability. The final injury is determined based on the injury code and injury notification probability transmitted from the module 230 and the injury selection module 250 and output (step S70).

For example, the injury final judgment module 260 determines that the injury code transmitted from the injury disease notification probability granting module 230 is a severe injury, that is, if the injury disease notification probability is 100%, it determines it as the final injury and outputs it, and the injury disease selection module 250 ) If the injury notification probability for each injury code transmitted from ) exceeds the preset standard value, it is judged as the final injury and output.

In some cases, the injury disease final determination module 260 sets the injury disease notification probability of the injury code transmitted from the injury selection module 250 to a specified value that matches the preset loss ratio (45%) as shown in FIG. If it exceeds (60% or 40%), it may be judged as a final corporal and printed. In this case, the regulation value matching the loss rate may be an injury disease notification probability value determined by considering the first year loss rate for each injury or disease.

As a way to run a stable business like this, the initial loss rate for each injury and disease can be set to n%, and the final injury rate can be determined by setting the injury notification probability that can converge to m% from a mid- to long-term perspective.

As described above, the injury/disease determination system 200 according to an embodiment of the present invention collects injury-related thinking ability information related to the insured person and generates thinking ability information for each disease, so the information on the injury codes included in the prescription is related to thinking ability. Not only can it be included in the information, but the injury notification probability is predicted for all injury codes included in the thinking ability information, and injury selection is performed by analyzing the correlation between each predicted injury code, so it is possible to identify diseases that must be preceded when screening for personal insurance subscription. There is an advantage in providing an injury/disease judgment system that makes automatic judgments, minimizes errors, and guarantees reliability.

Furthermore, the injury and disease determination system 200 according to an embodiment of the present invention has the advantage of reducing the workload of the reviewer because it automatically determines and notifies the injury or disease that requires review by the reviewer.

In the above, the present invention has been described with specific details such as specific components and limited embodiments and drawings, but this is only provided to facilitate a more general understanding of the present invention, and the present invention is not limited to the above embodiments. , a person skilled in the art to which the present invention pertains can make various modifications and variations from this description. Therefore, the spirit of the present invention should not be limited to the above-described embodiments, and the scope of the patent claims described below as well as all modifications equivalent to or equivalent to the scope of the claims fall within the scope of the spirit of the present invention. They will say they do it.

Claims (1)

In an injury judgment system executable on a computer system,
a thinking ability generation module that collects thinking ability information of the insured person listed in notification history information received online and generates thinking ability information for each injury or disease;
A thinking ability classification module that classifies the transmission path of the thinking ability information for each disease according to a plurality of classification criteria;
an injury notification probability granting module that grants an injury notification probability for each injury code by applying the thinking ability information for each injury transmitted according to the classification criteria to a predefined rule;
An injury notification probability prediction module that predicts the injury notification probability for each injury code by applying the thinking ability information for each injury delivered according to the classification criteria to a logistic regression model;
an injury selection module that selects an injury code to be notified based on a correlation analysis result of each injury code for which the injury notification probability is predicted;
An injury and disease determination system comprising: an injury and disease final determination module that determines and outputs a final injury based on the injury code and injury notification probability transmitted from the injury and disease notification probability granting module and the injury and disease selection module.