CN106447363A - Method and device for mining user category information - Google Patents

Abstract

The invention discloses a method and device for mining user category information. The method includes: for each application downloaded by a user in a user sample set whose user category information is known, labeling the application with user category information according to the number of downloads of the application by different types of users in the user sample set ; For a specified user whose user category information is to be mined, classify the applications downloaded by the specified user according to the marked user category information; for each user category, calculate the specified user category according to the relevant parameters of the application belonging to the user category Score value in the user category; determine the user category information of the designated user as being the same as the user category with the largest score value. The technical scheme of the present invention has the beneficial effect of reliably mining user category information based on the obtained data samples.

Description

A method and device for mining user category information

technical field

The invention relates to the field of data mining, in particular to a method and device for mining user category information.

Background technique

Targeted advertising is the precise delivery method in the advertising delivery strategy. Different advertisements are placed according to different target groups. When the target group meets the specified targeting characteristics, the advertisement is delivered to the corresponding user group, which can greatly improve the precise delivery of the advertisement and increase the Great advertising ROI. The judgment of user category information is the core of the problem. The traditional method of mining user category samples is to obtain user category information through user registration, questionnaire survey, etc. However, in this way, because most users do not fill in relevant information, and the information filled in is wrong, etc., the amount of information obtained is small, and the accuracy rate is not high.

Contents of the invention

In view of the above problems, the present invention is proposed to provide a method for mining user category information that overcomes the above problems or at least partially solves the above problems.

According to one aspect of the present invention, a method for mining user category information is provided, including:

For each of the applications downloaded by users in the user sample set whose user category information is known, mark the application with user category information according to the number of downloads of the application by different types of users in the user sample set;

For a specified user whose user category information is to be mined, classify each application downloaded by the specified user according to the marked user category information;

For each user category, calculate the score value of the specified user in the user category according to the relevant parameters of the application belonging to the user category;

The user category information of the designated user is determined to be the same as the user category with the largest score value.

Optionally, for each of the applications downloaded by users in the user sample set whose user category information is known, mark the application with user category information according to the number of downloads of the application by different types of users in the user sample set include:

For each application described, calculate the average number of downloads of the application among users in each user category;

Label the app with the same user category information as the user category with the largest average downloads.

Optionally, for each user category, according to relevant parameters of applications belonging to the user category, calculating the score value of the specified user in the user category includes:

For each user category, calculate the score value of the specified user in the user category according to the chi-square score and the chi-square score weight of each application belonging to the category.

Optionally, for each user category, calculating the score value of the specified user in the user category according to the chi-square score and the chi-square score weight of each application belonging to the category includes:

Calculate the score value S(u) _L of the specified user u in the user category L according to the following formula:

Among them, s _i is the chi-square score of the i-th application downloaded by the specified user, w _i is the chi-square score weight of the _i -th application downloaded by the specified user, and Li is the i-th application downloaded by the specified user The user category to which the app belongs.

Optionally, when there are only two user categories, the chi-square score s of an application is calculated according to the following formula:

s＝(A+B+C+D)(A*DB*C) ² /((A+C)*(B+D)*(A+B)*(C+D)).

Among them, A represents the number of downloads of the application by users belonging to the first user category in the user sample set; B represents the number of downloads of the application by users belonging to the second user category in the user sample set; C represents the number of downloads of the application in the user sample set. The number of times that users of the first user category do not download the application, and D represents the number of times that users belonging to the second user category in the user sample set do not download the application.

Optionally, when there are only two user categories, the chi-square score weight w of an application is equal to the ratio of the larger average download volume to the smaller average download volume of the application in the two user categories, specifically calculated according to the following formula An applied chi-square score weight w:

w=round(max(b(L))/min(b(L)))

where b(L) is an average download for users in user category L.

Optionally, the method further includes:

After the user category information of the designated user is determined, the designated user is added to the user sample set.

Optionally, the method further includes:

When the number of users added to the user sample set reaches a preset value, re-mark user category information on each application downloaded by the user in the user sample set.

Optionally, the method further includes:

For the designated user, push recommended content to the designated user according to the determined user category information.

According to another aspect of the present invention, a device for mining user category information is provided, wherein the device includes:

The labeling unit is suitable for each application downloaded by a user in the user sample set whose user category information is known, and marks the user category for the application according to the number of downloads of the application by different types of users in the user sample set information;

The mining unit is adapted to classify each application downloaded by the specified user according to the marked user category information for a designated user whose user category information is to be mined; for each user category, according to the relevant parameters of the application belonging to the user category, Calculate the score value of the specified user in the user category; determine the user category information of the specified user to be the same as the user category with the largest score value.

Optionally, the identification unit is adapted to, for each application, calculate the average download volume of the application among users of each user category; mark the application with the same user category as the user with the largest average download volume category information.

Optionally, the mining unit is adapted to, for each user category, calculate the score value of the specified user in the user category according to the chi-square score and the chi-square score weight of each application belonging to the category.

Optionally, the mining unit is adapted to calculate the score value S(u) _L of the specified user u in the user category L according to the following formula:

Among them, s _i is the chi-square score of the i-th application downloaded by the specified user, w _i is the chi-square score weight of the _i -th application downloaded by the specified user, and Li is the i-th application downloaded by the specified user The user category to which the app belongs.

Optionally, when there are only two types of users, the mining unit is adapted to calculate the chi-square score s of an application according to the following formula:

s＝(A+B+C+D)(A*DB*C) ² /((A+C)*(B+D)*(A+B)*(C+D)).

Among them, A represents the number of downloads of the application by users belonging to the first user category in the user sample set; B represents the number of downloads of the application by users belonging to the second user category in the user sample set; C represents the number of downloads of the application in the user sample set. The number of times that users of the first user category do not download the application, and D represents the number of times that users belonging to the second user category in the user sample set do not download the application.

Optionally, when there are only two user categories, the mining unit is adapted to set the chi-square score weight w of an application to be equal to the larger average download amount and the smaller average download amount of the application in the two user categories The ratio of downloads, more specifically, the chi-square score weight w of an application is calculated according to the following formula:

w=round(max(b(L))/min(b(L)))

where b(L) is an average download for users in user category L.

Optionally, the apparatus further includes: an adding unit, adapted to add the specified user to the user sample set after determining the user category information of the specified user.

Optionally, the identification unit is adapted to remark user category information on each application downloaded by a user in the user sample set when the number of users added to the user sample set reaches a preset value.

Optionally, the device further includes: a recommending unit, adapted to, for the specified user, push recommended content to the specified user according to the determined user category information.

According to the method for mining user category information of the present invention, for each application downloaded by the user in the collection of user samples with known user category information, the user of the application can be marked according to the number of times users of different categories download the application Category information, and for a designated user whose user category information is to be mined, classify the downloaded applications according to the marked user category information, and calculate the specified user’s status in this category according to the relevant parameters of the application belonging to the user category for each user category. The score value of the user is finally determined by the user category with the largest score value as the user category information of the specified user. This solves the problem of untruthful user category information obtained in the prior art, and achieves the beneficial effect of reliable mining of user category information based on the obtained data samples, thereby more accurately judging user category, and targeted advertising delivery, Greatly improve the precise placement of advertisements and increase the return on advertising investment.

The above description is only an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable , the specific embodiments of the present invention are enumerated below.

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiment. The drawings are only for the purpose of illustrating a preferred embodiment and are not to be considered as limiting the invention. Also throughout the drawings, the same reference numerals are used to designate the same components. In the attached picture:

Fig. 1 shows a flow chart of a method for mining user category information provided by an embodiment of the present invention;

Fig. 2 shows a structural diagram of an apparatus for mining user category information provided by an embodiment of the present invention.

detailed description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided for more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

Fig. 1 shows a flow chart of a method for mining user category information provided by an embodiment of the present invention. As shown in Fig. 1, the method includes:

Step S110, for each of the applications downloaded by users in the user sample set whose user category information is known, mark the application with user category information according to the number of downloads of the application by different types of users in the user sample set.

User category information can be divided according to the data obtained from the sample, such as workers, students, teachers, civil servants, etc. by occupation, or male and female by gender, or travel experts, shoppers, and technology experts by hobbies. Wait.

Step S120, for a designated user whose user type information is to be mined, classify the applications downloaded by the designated user according to the marked user type information.

Step S130, for each user category, calculate the score value of the specified user in the user category according to the relevant parameters of the applications belonging to the user category.

In step S140, the user category information of the specified user is determined to be the same as the user category with the largest score value.

For example, take user category information specifically referring to gender information as an example: a data sample set has been obtained, including 1,000 user samples with known gender, and the number of application software downloaded by each user, such as mobile APP, has been analyzed. Assume that the 1000 sample users have downloaded 1000 APPs in total. For each APP, count the number of downloads of the APP by users of different genders. If 200 users have downloaded it (the remaining 500 people have not downloaded it), then the user category information marked on the APP may be male (there are many ways to judge user category information, and here it is simply based on the number of users of each category in the sample judge by comparison); in analogy to this method, the remaining 999 APPs are counted on user gender information, that is, each APP is marked as male or female. Then, for a user whose gender information is to be mined, that is, to determine whether the user is male or female, the APPs downloaded by the user can be classified according to the corresponding labeling information of each APP; for example, the user has downloaded 40 APPs, and the 40 All APPs have been labeled with user classification information, of which 26 are labeled as male and 14 are labeled as female. At this time, for each type of APP, according to the relevant parameters of the application belonging to the user category, the score value of the specified user in the user category is calculated. Assuming that the parameters of each APP are recorded as 1, then the score of the user in the male category is 26. The score of the female category is 14, and the user category information of the specified user is determined to be the same as the user category with the largest score value, that is, the user is judged as a male.

This method is different from the questionnaire survey and user registration information in the traditional technology to judge the user category. The user category information obtained is not accurate enough, and it is difficult to collect a large number of samples and use machine learning to classify them. The method shown in Figure 1 can effectively solve this problem.

In yet another embodiment of the present invention, in the method shown in FIG. 1 , for each of the applications downloaded by users in the user sample set whose user category information is known, according to the different categories in the user sample set The number of times the user has downloaded the app, and the user category information for the app includes:

For each application, calculate the average download volume of the application among users of each user category; label the application with the same user category information as the user category with the largest average download volume.

In this embodiment, by calculating the average download volume of each application among users of each user category, the application is marked with the same user category information as the user category with the largest average download volume. Still taking the above-mentioned 1000 sample users with known genders as an example, for a review app (just for example), 300 male users downloaded the app, and 200 female users downloaded the app (the remaining 500 people did not download it, the 500 There are 200 male category users and 300 female category users), then for male category users, there are 500 user samples, and 300 people have downloaded the APP, then the average download volume is 300/500=0.6. Similarly, the average number of downloads among female category users is 0.4. If the comparison shows that the user category with the largest average download volume is male, then the user category information of the APP is marked as male. Analogous to this method, calculate the average downloads for the remaining apps and mark the user category information. This method can more accurately label user category information based on the relevant data of the application in the user sample.

In yet another embodiment of the present invention, in the method shown in FIG. 1, for each user category, calculating the score value of the specified user in the user category according to the relevant parameters of the application belonging to the user category includes:

For each user category, calculate the score value of the specified user in the user category according to the chi-square score and the chi-square score weight of each application belonging to the category.

Chi-square test is a hypothesis testing method with a wide range of uses. Its application in statistical inference of classified data includes: chi-square test for comparison of two rates or two constituent ratios; Chi-square test and correlation analysis of categorical data, etc. Only the chi-square score comparison is used in this implementation, and a more accurate result can be obtained by calculating the score value of the specified user in the user category by calculating the chi-square score.

In yet another embodiment of the present invention, in the above method, for each user category, the score value of the specified user in the user category is calculated according to the chi-square score and the chi-square score weight of each application belonging to the category include:

Calculate the score value S(u) _L of the specified user u in the user category L (when the user category specifically refers to the user's gender, L=male/female) according to the following formula:

Among them, s _i is the chi-square score of the i-th application downloaded by the specified user, w _i is the chi-square score weight of the _i -th application downloaded by the specified user, and Li is the i-th application downloaded by the specified user The user category to which the app belongs.

In yet another embodiment of the present invention, in the above method, when there are only two types of users, the chi-square score s of an application is calculated according to the following formula:

s＝(A+B+C+D)(A*DB*C) ² /((A+C)*(B+D)*(A+B)*(C+D)).

Among them, A represents the number of downloads of the application by users belonging to the first user category in the user sample set; B represents the number of downloads of the application by users belonging to the second user category in the user sample set; C represents the number of downloads of the application in the user sample set. The number of times that users of the first user category do not download the application, and D represents the number of times that users belonging to the second user category in the user sample set do not download the application.

This embodiment is suitable for judging only two types of user types, such as judging user gender, and the calculation is more convenient and fast.

In yet another embodiment of the present invention, in the above method, when there are only two user categories, the chi-square score weight w of an application is equal to the larger average download volume and the smaller average download volume of the application in the two user categories. The ratio of downloads, specifically calculate the chi-square score weight w of an application according to the following formula:

w=round(max(b(L))/min(b(L)))

Wherein, b(L) is an average download amount used among users of user category L, and round is a function of rounding the value according to a specified number of digits.

In yet another embodiment of the present invention, the above method further includes:

After the user category information of the designated user is determined, the designated user is added to the user sample set.

After the user category information of a user is specified, it can be added to the existing user sample set, which is conducive to expanding the user sample size and making the user information labeling based on data more accurate.

In yet another embodiment of the present invention, the above method further includes:

When the number of users added to the user sample set reaches a preset value, re-mark user category information on each application downloaded by the user in the user sample set.

In the above method, when too many newly added user samples reach the preset value, continuing the original method may cause a decline in the accuracy of user category information labeling. By re-labeling the user category of each application downloaded by the user in the user sample set Information can effectively address this issue.

In yet another embodiment of the present invention, the above method further includes:

For the designated user, push recommended content to the designated user according to the determined user category information.

This embodiment provides operations that can be performed after marking the user category information of the specified user. For advertising, it is of great significance to launch different advertisements according to different target groups. For example, if the user category is marked as a business person, advertisements for business supplies and suits can be placed; if the user category is a middle school student, teaching supplementary advertisements can be placed, and so on. Apps can also be recommended for people with different hobbies, such as recommending photography apps and photography magazine apps to photographers, and fashion magazine apps to makeup experts.

Fig. 2 shows a structural diagram of a device for mining user category information provided by an embodiment of the present invention. As shown in Fig. 2, the device 200 for mining user category information includes:

The labeling unit 210 is adapted to, for each application downloaded by a user in the user sample set whose user category information is known, mark the user for the application according to the number of downloads of the application by different types of users in the user sample set category information;

The mining unit 220 is adapted to, for a specified user whose user category information is to be mined, classify the applications downloaded by the specified user according to the marked user category information; for each user category, according to the relevant parameters of the application belonging to the user category , calculating the score value of the specified user in the user category; determining the user category information of the specified user to be the same as the user category with the largest score value.

It is different from the questionnaire survey and user registration information in the traditional technology to judge the user category, because in the above-mentioned traditional technology, many users do not provide real information when filling in, or for reasons such as protecting privacy, because they are too lazy to fill in. Provide information, the user category information obtained is not accurate enough, and it is difficult to collect a large number of samples and use machine learning to classify, but the device shown in Figure 2 can effectively solve this problem.

In yet another embodiment of the present invention, the identification unit 210 in FIG. 2 is adapted to, for each application, calculate the average download volume of the application among users of each user category; label the application with the average The same user category information as the most downloaded user category.

Using this method, the identification unit 210 can more accurately mark the user category information according to the relevant data applied in the user sample.

In yet another embodiment of the present invention, the mining unit 220 is adapted for each user category, according to the chi-square score and the chi-square score weight of each application belonging to the category, to calculate the specified user's The score value for the category.

In yet another embodiment of the present invention, the mining unit 220 in the above-mentioned device is adapted to calculate the score value S(u) _L of the specified user u in the user category L according to the following formula:

Among them, s _i is the chi-square score of the i-th application downloaded by the specified user, w _i is the chi-square score weight of the _i -th application downloaded by the specified user, and Li is the i-th application downloaded by the specified user The user category to which the app belongs.

The mining unit 220 calculates the score value of the specified user in the user category by calculating the chi-square score to obtain a more accurate result.

In yet another embodiment of the present invention, in the above device, when there are only two types of users, the mining unit 220 is adapted to calculate the chi-square score s of an application according to the following formula:

s＝(A+B+C+D)(A*DB*C) ² /((A+C)*(B+D)*(A+B)*(C+D)).

Among them, A represents the number of downloads of the application by users belonging to the first user category in the user sample set; B represents the number of downloads of the application by users belonging to the second user category in the user sample set; C represents the number of downloads of the application in the user sample set. The number of times that users of the first user category do not download the application, and D represents the number of times that users belonging to the second user category in the user sample set do not download the application.

This embodiment is suitable for judging only two types of user types, such as gender judgment, and the calculation is more convenient and quick.

In yet another embodiment of the present invention, in the above device, when there are only two user categories, the mining unit 220 is adapted to set the chi-square score weight w of an application to be equal to that of the application in the two user categories The ratio of the larger average download volume to the smaller average download volume in , more specifically, calculate the chi-square score weight w of an application according to the following formula:

w=round(max(b(L))/min(b(L)))

where b(L) is an average download for users in user category L.

In yet another embodiment of the present invention, the above-mentioned device further includes:

The adding unit is adapted to add the specified user to the user sample set after determining the user category information of the specified user.

After the user category information of a user is specified, it can be added to the existing user sample set by using the adding unit, which is conducive to expanding the user sample capacity and making the user information labeling based on data more accurate.

In yet another embodiment of the present invention, the identification unit in the above-mentioned device is adapted to re-identify each user downloaded by the user in the user sample set when the number of users added to the user sample set reaches a preset value. App label user category information.

When the adding unit adopts the above method to add too many user samples and reaches the preset value, continuing the original method may cause a decline in the accuracy of user category information labeling. By re-labeling each application downloaded by the user in the user sample set User category information can effectively solve this problem.

In yet another embodiment of the present invention, the above-mentioned device further includes:

The recommending unit is adapted to, for the specified user, push recommended content to the specified user according to the determined user category information.

This embodiment provides operations that can be performed after marking the user category information of the designated user. For advertising, it is of great significance to launch different advertisements according to different target groups. For example, if the user category is marked as a business person, advertisements for business supplies and suits can be placed; if the user category is a middle school student, teaching supplementary advertisements can be placed, and so on. Apps can also be recommended for people with different hobbies, such as recommending photography apps and photography magazine apps to photographers, and fashion magazine apps to makeup experts.

In a specific embodiment of the present invention, a batch of user sample sets whose user category information is known is obtained, wherein user categories are divided into male and female, 55376 male samples and 52902 female samples. First calculate the average downloads of the APP in the male category and the female category, according to the formula

L = male, female

The calculation results are obtained, where appL represents the number of downloads of the APP under category L, C _L represents the total number of samples of sample L, and then according to the formula

L ^* ＝argmax _L {b(L)}, L＝male, female

Determine the user category of the APP. Then take the ratio of the larger average download volume to the smaller average download volume of the APP in the two categories as the weight of the APP chi-square value, that is, use the aforementioned formula

w=round(max(b(L))/min(b(L)))

get. When the total number of downloads of the APP in the two categories is less than 10, set w to 1. If the APP only appears in one category, use the formula

w=round(count/10)

Calculate, where count is the number of times the app has been downloaded in this category. Then according to the above formula

Get the score value S(u) _L of the specified user u in the user category L, then the user category is the category with the largest category score. The result is as follows:

AUC＝0.782, total correct rate＝75.8%

It can be seen that the total accuracy rate of the user test set sample is above 75%.

In addition, when there are a large number of labeled samples, samples with higher quality can be filtered out as the user sample set, and only when the labeled samples are consistent with the calculated user category will be screened out.

When the sample size is insufficient, the user type of the user can be directly mined through the technical solution provided by the present invention.

It should be noted:

The algorithms and displays presented herein are not inherently related to any particular computer, virtual appliance, or other device. Various general purpose devices can also be used with the teachings based on this. The structure required to construct such an apparatus will be apparent from the foregoing description. Furthermore, the present invention is not specific to any particular programming language. It should be understood that various programming languages can be used to implement the content of the present invention described herein, and the above description of specific languages is for disclosing the best mode of the present invention.

In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure the understanding of this description.

Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, in order to streamline this disclosure and to facilitate an understanding of one or more of the various inventive aspects, various features of the invention are sometimes grouped together in a single embodiment, figure, or its description. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this invention.

Those skilled in the art can understand that the modules in the device in the embodiment can be adaptively changed and arranged in one or more devices different from the embodiment. Modules or units or components in the embodiments may be combined into one module or unit or component, and furthermore may be divided into a plurality of sub-modules or sub-units or sub-assemblies. All features disclosed in this specification (including accompanying claims, abstract and drawings) and any method or method so disclosed may be used in any combination, except that at least some of such features and/or processes or units are mutually exclusive. All processes or units of equipment are combined. Each feature disclosed in this specification (including accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will understand that although some embodiments described herein include some features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of the invention. and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all functions of some or all components in the device for mining user category information according to the embodiment of the present invention. The present invention can also be implemented as an apparatus or an apparatus program (for example, a computer program and a computer program product) for performing a part or all of the methods described herein. Such a program for realizing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet site, or provided on a carrier signal, or provided in any other form.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means can be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. does not indicate any order. These words can be interpreted as names.

Claims (10)

1. a kind of method of digging user classification information, wherein, the method includes：

Every in each application that user in user's sample set known to user category information is downloaded Individual application, downloads the number of times of this application according to the different classes of user in user's sample set, for this application Mark user category information；

For the specified user that a user category information is to be excavated, each application that this specified user is downloaded User category information according to mark is classified；

For each class of subscriber, according to the relevant parameter of the application belonging to this class of subscriber, calculate this and refer to Determine the fractional value in this class of subscriber for the user；

The user category information of this specified user is defined as the class of subscriber maximum with fractional value identical.

2. the method for claim 1, wherein for user's sample known to user category information Each application in each application that user in set is downloaded, according to the inhomogeneity in user's sample set Other user downloads the number of times of this application, includes for this application mark user category information：

For each described application, calculate this and apply the average download in the user of each class of subscriber Amount；

For the class of subscriber identical user category information that this application mark is maximum with average download.

3. the method for claim 1, wherein for each class of subscriber, according to belonging to this use The relevant parameter of the application of family classification, calculates this specified user and includes in the fractional value of this class of subscriber：

For each class of subscriber, the card side's score value according to each application belonging to the category and Ka Fang score value are weighed Weight, calculates the fractional value in this class of subscriber for this specified user.

4. method as claimed in claim 3, wherein, for each class of subscriber, according to belonging to such Card side's score value of other each application and Ka Fang score value weight, calculate this specified user dividing in this class of subscriber Numerical value includes：

Calculate fractional value S (u) specifying user u in class of subscriber L according to equation below_L：

$S{\left(u\right)}_L=\underset{L_i\∈L}{\Σ}\log (s_i+1)*\log \left(w_i\right)$

Wherein, S_iIt is card side's score value of i-th application that this specified user downloads, W_iIt is this specified user Card side's score value weight of i-th application downloaded, L_iIt is belonging to i-th application that this specified user downloads Class of subscriber.

5. method as claimed in claim 4, wherein, when only two kinds of class of subscribers, according to as follows Formula calculates the card side score value s of an application：

S=(A+B+C+D) (A*D-B*C)²/((A+C)*(B+D)*(A+B)*(C+D))

Wherein, A represents that the user belonging to first user classification in user's sample set downloads this application Number of times, B represents that the user belonging to second user classification in user's sample set downloads the number of times of this application； C represents that the user belonging to first user classification in user's sample set does not download the number of times of this application, D Represent that the user belonging to second user classification in user's sample set does not download the number of times of this application.

6. method as claimed in claim 4, wherein, when only two kinds of class of subscribers, one should Card side's score value weight w equal to this apply larger average download in two class of subscribers with less The ratio of average download, calculates card side's score value weight w of an application with specific reference to equation below：

W=round (max (b (L))/min (b (L)))

Wherein, b (L) is an average download being used in the user that user classifies L.

7. the method as any one of claim 1-6, wherein, the method further includes：

After determining the user category information of this specified user, this specified user is added to described user's sample In set.

8. method as claimed in claim 7, wherein, the method further includes：

When the number of users being added in described user's sample set reaches preset value, again to user's sample Each application mark user category information that user in this set is downloaded.

9. the method as any one of claim 1-6, wherein, the method further includes：

For this specified user, according to determined by user category information pushes to this specified user and recommends Hold.

10. a kind of device of digging user classification information, wherein, this device includes：

Mark unit, is suitable to be downloaded for the user in user's sample set known to user category information Each application in each application, this application is downloaded according to the different classes of user in user's sample set Number of times, marks user category information for this application；

Excavate unit, be suitable to the specified user to be excavated for a user category information, this is specified and uses Each application that family is downloaded is classified according to the user category information of mark；For each class of subscriber, root According to the relevant parameter of the application belonging to this class of subscriber, calculate the fraction in this class of subscriber for this specified user Value；The user category information of this specified user is defined as the class of subscriber maximum with fractional value identical.