JP6929698B2 - Information processing equipment, information processing methods, and information processing programs - Google Patents

Description

The present invention relates to an information processing device, an information processing method, and an information processing program for classifying a plurality of classification objects.

Conventionally, there is known a technique for grouping a plurality of classification objects according to the attributes of the classification object (see, for example, Patent Document 1).
In the advertisement distribution server of Patent Document 1, the attribute information of the user is recorded in the user attribute database, and the user attribute analysis means analyzes the attribute information to derive the factor attribute which is an objective attribute of the user. This factor attribute is a category score in the population, and is a value indicating how positive the category is. Then, when the advertisement is distributed, the advertisement distribution server extracts user data having a factor attribute corresponding to the advertisement attribute (keyword group) and targets the advertisement distribution target.
That is, the advertisement distribution server described in Patent Document 1 classifies a plurality of users into a plurality of groups, with users having the same factor attribute (classification target) as one group, and advertisements corresponding to each group. Deliver to users who belong to the group.

JP-A-2007-26456

By the way, in the prior art as described in Patent Document 1, when the classification objects are grouped according to the attributes, the number of classification objects belonging to each group is not necessarily the number of classification objects targeted by the group. May not be.
For example, when dividing a user into a plurality of groups in marketing activities, the users are divided into groups under the following conditions.
(A) One user can belong to a plurality of groups.
(B) There is an upper limit to the number of groups to which one user belongs.
(C) There is a target value for the number of users belonging to each group.
In this case, conventionally, after calculating the score for each group for each user by using a technique as shown in Patent Document 1, for example, the group having a large score is selected by the upper limit number (hereinafter, G). , That group was the group to which the user belongs. Here, if the number of users belonging to each group is close to the target value, there is no problem. However, in reality, the number of users belonging to the group often deviates from the target value.
To give a specific example, when a user is divided into a male group and a female group according to registered attribute information, behavior history, etc., the number of users in the male group tends to be larger than the number of users in the female group. be. However, in the actual market, there is greater demand for sales promotion activities for women than for men. Advertisers want to serve as many female users as possible when delivering ads for women's products, but as mentioned above, when the users are grouped, the number of users belonging to the female group is small. In that case, the target value of distribution and the number of users deviate from each other, and the advertising effect expected by the advertiser cannot be sufficiently obtained.

An object of the present invention is to provide an information processing device, an information processing method, and an information processing program that can appropriately classify a classification target with respect to a target value of each group.

The information processing apparatus of the present invention includes an attribute acquisition unit that acquires attribute information possessed by the classification target and a group setting unit that divides the classification target into groups, and the group setting unit relates the classification target to the group. When the classification target is grouped using the score indicating the degree, the score is corrected and the number of the classification target belonging to the group is adjusted based on the information of the classification target.

In the present invention, the score is corrected based on the information of the classification target to adjust the number of classification targets belonging to the group. As a result, the number of classification targets belonging to each group can be brought close to the target value, and each classification target can be appropriately classified.

The block diagram which shows the schematic structure of the information processing system of 1st Embodiment. The block diagram which shows the schematic structure of the server apparatus of 1st Embodiment. The block diagram which shows the schematic structure of the terminal apparatus. The flowchart which shows the information processing method of 1st Embodiment. The flowchart which shows the detail of the correction value correction processing in 2nd Embodiment. The figure which shows an example of the transition of the ratio of the number of users and a target value in 2nd Embodiment.

[First Embodiment]
Hereinafter, the information processing apparatus of the first embodiment according to the present invention will be described.
FIG. 1 is a block diagram showing a schematic configuration of an information processing system 1 including a server device 10 which is an information processing device according to the present invention.
As shown in FIG. 1, the information processing system 1 is constructed by a server device 10 and a plurality of terminal devices 20 that are communicably connected to the server device 10 via a network (for example, the Internet or the like).

In the information processing system 1, the server device 10 classifies the users who operate the terminal device 20 (classification targets in the present embodiment) into a plurality of groups based on various information transmitted from the terminal device 20.
Here, an example of classifying users by the server device 10 is shown as an example, but the server device 10 may perform various processes corresponding to each group for the users classified into each group. good. For example, the server device 10 may deliver contents such as advertisements according to the attributes of each group to users belonging to each group, and provides specific information about users belonging to each group to affiliated companies. You may.

[Configuration of server device 10]
FIG. 2 is a block diagram showing a schematic configuration of the server device 10 of the present embodiment.
The server device 10 of the present embodiment is composed of a computer, and includes a communication unit 11, a recording unit 12, a control unit 13, and the like.
The communication unit 11 is connected to the network and communicates with the terminal device 20 and other devices on the network via the network.

The recording unit 12 is a data recording device composed of, for example, a memory, a hard disk, or the like. The recording unit 12 includes a user information recording area (user DB 12A) for recording various data related to users to be classified according to the present invention, and a group information recording area (user DB 12A) for recording various information related to each group for grouping users. Group DB12B) and the like are provided.

The user DB 12A is a database in which a plurality of user information in which a plurality of attribute data indicating the characteristics of each user are recorded is recorded. This user information includes, for example, user identification information (user ID), user attribute information, affiliation group information, and the like.
The user identification information is information for identifying each user, and has different information for each user.
The user attribute information is information indicating the characteristics of the user, and includes various information about the user such as gender, age or age group, whereabouts, occupation, hobbies, fields of interest, and the like. Further, the user attribute information includes input information input by the user operating the terminal device 20 (for example, a keyword input by user registration or questionnaire response), as well as information detected by the user's behavior. May be good.
The information detected by the user's behavior includes, for example, the operation history of the user's terminal device 20 for each predetermined period and the behavior history on the Internet for each predetermined period (for example, the history of keywords in the search process and the history of browsing pages). , Purchase history of products, download history, etc.). Further, a sensor that detects specific information in the terminal device 20 (for example, a GPS device that detects current position information, an acceleration sensor or motion sensor that detects movement or posture such as acceleration, a temperature sensor that detects temperature, or blood flow. If a sensor that detects health information such as blood pressure or blood pressure is installed, the detection value (sensor detection history) for each predetermined period detected by the sensor may be recorded.
The affiliation group information is information about the group to which the user belongs, and for example, the group ID of the group information described later is recorded.

The group DB 12B is a database in which a plurality of group information in which information for managing each group and information indicating the characteristics of each group are recorded is recorded. This group information includes, for example, group identification information (group ID), target value, score calculation method, and the like.
The group identification information is information for identifying each group, and each group has different information.
The target value is a target value for the number of users belonging to the group, and is set for each group.
The score calculation method is a rule (mathematical formula, logical formula, etc.) for calculating the score for each group based on the user attribute information. The group information includes other group names, services provided to users belonging to the group, and distributed contents. Information about the recording location of the content, the company that provides the service to the users belonging to the group, and the like may be associated with the information. Further, in the present embodiment, an example of recording the score calculation method is shown, but the conditions of the users belonging to the group, the attributes of the group, and the like may be separately recorded.
Further, in the present embodiment, the user information includes the affiliation group information, and finally the information about the group to which the user belongs is recorded, but the group information includes the information about the user who belongs to the group (for example, the user ID). ) May be recorded.

Further, the recording unit 12 records a plurality of programs to be executed by the control unit 13. These programs include a program for executing an operating system that controls each part of the server device 10, various application programs executed on the operating system (information processing program (classification processing program) of the present invention, etc.).

The control unit 13 is composed of an arithmetic circuit such as a CPU (Central Processing Unit) and a recording circuit such as a RAM (Random Access Memory). The control unit 13 expands the program recorded in the recording unit 12 or the like into the RAM, and executes various processes in cooperation with the program expanded in the RAM. Then, the control unit 13 reads and executes the information processing program (classification processing program) recorded in the recording unit 12, and as shown in FIG. 2, the user information acquisition unit 131, the group information acquisition unit 132, and the group It functions as a setting unit 133 and the like.

The user information acquisition unit 131 also functions as an attribute acquisition unit of the present invention, acquires user information transmitted from the terminal device 20 via the network, and records it in the user DB 12A.
The group information acquisition unit 132 acquires group information and records it in the group DB 12B. As for the acquisition of group information, for example, the group information is acquired from a predetermined device (for example, an advertiser terminal managed by an advertiser) via a network. Alternatively, the group information acquisition unit 132 may acquire the group information transmitted from a predetermined terminal by the administrator of the server device 10 via the network, or the group information input by the input means connected to the server device 10. May be obtained.
Further, the group information acquisition unit 132 may acquire the above-mentioned group information, that is, the target value and the score calculation method, and the conditions of the users who belong to the group (for example, the conditions of the user who is the target of the distribution of the advertisement content desired by the company). , The type of advertising content, etc.) may be acquired. In this case, the group information acquisition unit 132 may set a target value and a score calculation method from the obtained information.

The group setting unit 133 classifies (groups) each user into each group and updates the group information to which the user information belongs.
Specifically, the group setting unit 133 includes the functional configurations of the score calculation unit 133A, the score correction unit 133B, the group selection unit 133C, and the group registration unit 133D.
The score calculation unit 133A calculates the score for each group in each user. This score is a value indicating the degree of association between the user and the group, and indicates how well the user meets the conditions of the group. In this embodiment, the higher the score, the more the user is associated with the group, that is, the closer the user's attributes are to the group's conditions, the higher the score value. As a method for calculating the score, a known technique (for example, Japanese Patent Application Laid-Open No. 2016-153972) can be used.

The score correction unit 133B corrects the score calculated by the score calculation unit 133A by using the correction value. In the present embodiment, the correction value is a value set for each group and is a value common to all users, but is increased or decreased depending on the relationship between the target value of the group and the number of users corresponding to the conditions of the group. NS.
The target value of the group is a target value of the number of users belonging to the group, and is set based on, for example, user attribute information. The number of users corresponding to the group condition is the number of users when the users are actually grouped based on the user attribute information. Therefore, it can be said that the score correction unit 133B corrects the score based on the user's information.

The group selection unit 133C selects a group corresponding to each user based on the score calculated by the score calculation unit 133A or the score corrected by the score correction unit 133B (correction score). Here, the group selected is a temporary group for the user, not a formally determined group.
When each user is divided into groups, the group registration unit 133D assigns the user to the group when the number of users belonging to each group becomes the target value set for each group or a value close to the target value. That is, the group ID of the selected group is recorded in the user information to which the user information belongs.

[Configuration of terminal device 20]
FIG. 3 is a block diagram showing a schematic configuration of a general terminal device 20.
As described above, the terminal device 20 is a computer owned by the user to be classified, and is composed of, for example, a smartphone, a tablet terminal, a personal computer, or the like. As shown in FIG. 3, the terminal device 20 includes, for example, a display unit 21, an input operation unit 22, a terminal communication unit 23, a terminal recording unit 24, a terminal control unit 25, and the like.
The display unit 21 is composed of, for example, a liquid crystal display or the like, and displays a predetermined image under the control of the terminal control unit 25.
The input operation unit 22 may be composed of, for example, a touch panel provided integrally with the display unit 21, or may be composed of an input device such as a keyboard or a mouse. The input operation unit 22 is operated by the user to output an operation signal corresponding to the operation to the terminal control unit 25.
The terminal communication unit 23 communicates with the server device 10 and a predetermined device on the network.

The terminal recording unit 24 is composed of a data recording device such as a memory or a hard disk. In the terminal recording unit 24, for example, in addition to user attribute information (for example, gender, age, whereabouts, field of interest) about the user input by the input operation of the input operation unit 22, when the user operates the terminal device 20 The operation history, the user's behavior history on the Internet, the detection result history detected by various sensors (for example, GPS device, motion sensor, etc.) provided in the terminal device 20 may be recorded. Further, various programs for controlling the terminal device 20 and the like are recorded in the terminal recording unit 24.

The terminal control unit 25 is composed of an arithmetic circuit such as a CPU (Central Processing Unit) and a recording circuit such as a RAM (Random Access Memory), and controls each part of the terminal device 20. The terminal control unit 25 expands the program (software) recorded in the terminal recording unit 24 or the like into the RAM, and executes various processes in cooperation with the program expanded in the RAM. Specifically, the terminal control unit 25 reads and executes the above program to activate a predetermined application program or activate various sensors of the terminal device 20 in response to the operation of the user's input operation unit 22. To acquire each sensing data. In addition, user attribute information, operation history, action history, and detection result history are acquired and recorded in the terminal recording unit 24. Further, the terminal control unit 25 transmits various information recorded in the terminal recording unit 24 to the server device 10 based on, for example, a user's operation.

[Operation of information processing system]
Next, the information processing method in the information processing system 1 in the present embodiment, in particular, the classification method of the users to be classified in the server device 10 will be described.
FIG. 4 is a flowchart showing an information processing method according to the present embodiment.
In the information processing system 1 of the present embodiment, the users to be classified are grouped (classified) into each group based on the user information and the group information. In particular, the information processing system 1 of the present embodiment is used for marketing of a company, for example, when a company that provides a service distributes contents such as advertisements and questionnaires of the service to be provided, or when a free sample such as a product is sent. It can be suitably used when selecting a target user in an activity.
For example, when an advertiser such as a company distributes advertisement contents to users via the Internet, the advertisement contents are distributed by selecting users based on the user information about each user held by the company. In this case, the company uses the server device 10 of the present embodiment to group the users into groups according to the service, and distributes the advertisement contents corresponding to each group to the users belonging to the group. Alternatively, for example, when requesting the distribution of advertising content to a site used by many users on the Internet, a so-called portal site, the server device 10 of the present embodiment is used. As a result, the server device 10 of the portal site divides the users into groups according to the services requested by each company, and distributes the advertisement contents corresponding to each group to the users belonging to the group.
In the following, as an example of the information processing method of the present embodiment, an advertiser such as a company sends an advertisement request to, for example, a server device 10 owned by a portal site or the like, and the server device 10 is a group in which the user corresponds to the advertisement. An example of classifying into is described.

In addition, in the grouping of users in this embodiment, the following three conditions (A) to (C), that is,
(A) One user can belong to a plurality of groups.
(B) There is an upper limit to the number of groups to which one user belongs.
(C) The number of users belonging to each group has a target value (target value of the number of users belonging to the group), and the closer to the target value, the better.
It is carried out based on.
Here, in the following description, the upper limit of the number of groups to which one user can belong is G, the number of users belonging to the group g (or temporary group g) is _Ug, and the target value of the number of users in the group. _{Let Ugt} .

In the information processing system 1 of the present embodiment, when the server device 10 groups each user into a plurality of groups, the user information acquisition unit 131 first acquires the user information of each user (step S1) and groups. The information acquisition unit 132 acquires the group information of each group (step S2).

In step S1, the user information acquisition unit 131 acquires user information from the terminal device 20 owned by each user.
For example, when the user operates the input operation unit 22 of the terminal device 20, the terminal device 20 accesses a predetermined content (for example, a user registration page) that can be viewed on the network. As a result, the user registration page is displayed on the display unit 21 of the terminal device 20. When the user inputs various user attributes (for example, gender, age, whereabouts, occupation, etc.) according to the guidance on the user registration page, the terminal device 20 transmits the user attribute information to the server device 10. At this time, when the user ID is input from the terminal device 20, the user information acquisition unit 131 identifies the user information from the user DB 12A and updates the user attribute information of the user information. If there is no user ID, new user information is generated, a new user ID is assigned, and the input user attribute information is recorded.
Further, the user information acquisition unit 131 may acquire user attribute information from the terminal device 20 periodically or at a predetermined timing set by the user, and update the corresponding user information of the user DB 12A. Specifically, the user information acquisition unit 131 receives from the terminal device 20 an operation history recorded in the terminal recording unit 24 of the terminal device 20, an action history on the Internet, a sensor detection history, and the like.

In step S2, the group information acquisition unit 132 generates or updates each group information.
In the present embodiment, group information regarding the distribution of advertising content is acquired from the advertiser. In this case, the group information acquisition unit 132 acquires the number of advertisement contents delivered and the score calculation method from the advertiser terminal owned by the advertiser.
Further, instead of the score calculation method, the conditions for the distribution target of the advertisement content may be acquired. In this case, the group information acquisition unit 132 calculates the score based on the condition of the distribution target of the advertisement content, that is, the condition of the target user belonging to the group.

As a score calculation method, for example, model generation based on correct answer information can be used.
To explain with a specific example, in the case of calculating the score related to the distribution of advertising content related to "sales of A cosmetics", the user who actually purchased "A cosmetics" is specified, and the user information of the user is correctly answered from the user DB 12A. Extract as information. In this case, the server device 10 transmits questionnaire information regarding "purchase of cosmetics A" to the terminal device 20, and when the answer of "purchased" is obtained from the terminal device 20, the user information of the user is correctly answered. May be. Further, in the user attribute information (for example, purchase history on the Internet), the user information in which "A cosmetics purchase" is recorded may be extracted as correct answer information. Then, by totaling and analyzing each user attribute recorded in a plurality of correct answer information (user information), a coefficient (weight value) for each user attribute is calculated. That is, among the users corresponding to the correct answer information, the weight value is increased for the user attribute common to many users, the weight value is decreased for the user attribute not common, and the score is calculated using the weight value as the score calculation coefficient. The weight value for each user attribute in each group may be adjusted by machine learning (neural network using deep learning, etc.).

Further, the group information acquisition unit 132 may appropriately set a target value based on information on the Internet or the like. For example, from research sites published on the Internet, articles on blogs, etc., public information on the Internet such as posting comments, the content of public information related to the service to be advertised (for example, "A cosmetics", etc.), public information Extract the number of cases. Then, based on the number of public information for the Internet population, the demand in the market may be predicted and the target value may be set. For example, a target value according to the demand in the market may be held as table data, and the target value may be set based on the table data.

Further, in FIG. 4, step S2 is performed after step S1, but the present invention is not limited to this. That is, the timing of user registration by the user, the transmission of user attribute information, and the registration of group information are carried out by the user or the advertiser at an arbitrary timing, and steps S1 and S2 may be carried out at the same time. , Step S1 may be carried out after step S2.

Then, the server device 10 determines whether or not to perform the user classification process (grouping) (step S3). That is, the server device 10 determines whether or not it is time to execute the classification process.
This timing may be, for example, a preset date and time, or may be every preset cycle. Further, it may be after the information is updated by step S1 or step S2. The timing after the information is updated by step S1 or step S2 may be, for example, immediately after the execution of step S1 or step S2, or after the execution of step S1 or step S2 and after the lapse of a preset time. There may be. Alternatively, the classification process may be performed at the timing of delivering the advertisement content to each user.
If No is determined in step S3, the process returns to step S1 and waits until the timing.

If it is determined Yes in step S3, the score calculation unit 133A of the server device 10, a score _{S ug} for the group g of user u, is calculated for all users and all groups (Step S4).
As a method for calculating the score, a known technique can be used as described above. For example, for a group, N values (attribute _values) of the individual user attributes _{A 1, A 2, A 3} , ... A i, ... A N-1, with respect to _{A N,} respectively weights _{k 1} , K ₂ , k ₃ , ... k _i , ... k _N-1 , k _N are set, the score calculation unit 133A calculates the score _Sug by the following equation (1).

Each value A _i of the user attributes are different for each user, since the score weight value k _i different for each group, the score S _ug is a different value for each group for each user.
When each value of the user attribute is a selection item such as gender, the value of the attribute may be appropriately modified according to each group. For example, in the score calculation for a group for "female cosmetics", the attribute value is "0" when the user attribute related to gender is "male", and the attribute value is "1" when the user attribute is "female", and the score _Sug is set. calculate. Further, in the user attribute related to the whereabouts, for example, the value may be increased when the distance is close to the target position of the group, and may be decreased when the distance is long. For example, in the score calculation for a group living in "Tokyo", if the user attribute related to the whereabouts is "Tokyo", the attribute value is "1.0", and in "Yamanashi prefecture", which is a neighboring prefecture of Tokyo, the attribute value is Is set to "0.8", and the attribute value is set to "0.5" in "Aichi prefecture" far from Tokyo.

The score calculation method is not limited to this, and the score may be calculated by another method. _{For example, a score Sug} for each group of each user may be calculated by using a machine learning model (for example, a neural network or the like) in which each user attribute is an input value and a score for each group is an output value.

Thereafter, the score correction unit 133B, the score calculated in step S4, then corrected by the correction value _{b g,} calculates the correction score _{T ug} (step S5).
Specifically, the score correction unit 133B calculates the _{correction score Tug} based on the following formula (2).

In the formula (2), the correction value b _g of the denominator is a value set for each group is independent of the user. In the present embodiment, the initial value of the correction values b _g is "1". _{Therefore, the correction score Tug} calculated in the first step S5 has the same value _{as the score Sug} calculated in step S4.

After that, the group selection unit 133C selects _{G groups for each user in descending order of the correction score Tug} calculated in step S5 (step S6). That is, only the upper limit G of the number of groups to which the user can belong is selected. The group selected here is not a group actually registered in the user information as a group to which the user belongs, but a temporary group.

Next, the group setting unit 133 determines the number of users U _g in each group to which the user is classified as a temporary group, the ratio between the target value U _gt of the group, whether falls within a predetermined range. Specifically, whether or not the condition shown in the following equation (3) is satisfied, that is, the ratio (U _g / U _{gt ) of the number of users U g} to the target value U _gt is within a predetermined range (from the lower limit threshold value P1). (Up to the upper limit threshold value P2) is determined (step S7).

In the equation (3), the threshold values P1 and P2 are values for determining whether or not the _{number of users U g} is close to the target value U _gt. Thus, for example, as the number of users _{U g} belonging to the group, to set the range of 80% of the target value _{U gt} of 110%, P1 = 0.8, may be set P2 = 1.1. Further, when it is strictly _{matched with the target value Ugt} , P1 = P2 = 1.0 may be set.
These threshold values P1 and P2 may be common values for each group, but different threshold values P1 and P2 may be set for each group. For example, the threshold values P1 and P2 may be set closer to 1 as _{the target value Ugt is smaller.}
If No is determined in step S7, it means that the number of users U _g and the target value U _gt deviate from each other. In this case, the score correction unit 133B, to correct the correction value _{b g} (step S8), and returns to step S5.
In step S8, the score correction unit 133B on the basis of the following equation (4), corrects the correction value _{b g.}

Here, α (where α> 0) is an index of the correction coefficient, and a fixed value is used in this embodiment.
In the present embodiment, as shown in the equation (4), the correction value b _{g used for calculating the correction score Tug} in the previous step S5 is multiplied by (U _g / U _gt ) ^α to be newly added. The correction value is b _g .
That is, since b _g = 1 in the first _{loop, the correction score Tug} = score _Sug , and _{grouping based on the score Sug} (selection of a provisional group) is performed. Then, when (U _g / U _gt ) does not fall within the predetermined range ( _{the values of U g} and U _gt deviate from each other), the correction value b _g is expressed by the number of users U as shown in the equation (4). Correct according to the ratio of _g to the target value U _{gt (U g} / U _gt).

In the correction of the correction value b _g by the formula (4), when the number of users U _g is larger than the target value U _gt , the correction value b _g becomes large, and the correction score T _{ug calculated by the formula (2).} Becomes smaller. When the number of users U _g is smaller than the target value U _gt , the correction value b _g becomes smaller and the correction score T _ug becomes larger.
The correction value _bg is set for each of the groups. That is, when the _{value of the number of users U g} and the target value U _gt deviate in the A group and the number of users U _g is close to the target value U _{gt in the B group, only the correction value b g} for the A group is corrected. Will be done. As a result, by executing the loop processing as described above a plurality of times, the number of users U _g belonging to each of all the groups is adjusted to approach the target value U _{gt value of each group.}

Then, in step S7, when it is determined that the above equation (3) is satisfied for all the groups (determined as Yes in step S7), the group registration unit 133D selects the temporary group selected for each user. Set as the group to which the user belongs (step S9). In this way, the users belonging to each group are determined, and the group registration unit 133D records and updates, for example, the group ID of the group to which the user belongs with respect to the group information to which the user belongs.
In this embodiment, an example of updating the user information is shown, but the user group information associated with each group and the user belonging to the group may be generated and recorded in the recording unit 12. Further, as described above, for example, a user ID of a user belonging to the group may be recorded in each group information.

After the classification process as described above, the server device 10 performs various processes corresponding to the group for the user (terminal device 20) belonging to each group. For example, the advertisement processing associated with the group is transmitted to the terminal device 20 of the user belonging to the group, or the questionnaire information is distributed.

[Action and effect of the first embodiment]
In the server device 10 of the present embodiment, the group setting unit 133 _{calculates the score Sug for the user's group based on the user information, corrects the score Sug} based on the user's information, and the number of users U belonging to the group. Adjust _g.
Thus, the score S _ug, since the number of users U _g is corrected so as to approach the target value U _gt, it is possible to properly classify the user for each group.

The above-mentioned action and effect will be described with reference to specific examples. When delivering advertising content for "cosmetics for women in their 50s," which is in high demand in the market, the conventional method (when not corrected by the correction value) adjusts the user to, for example, "gender,""age,""field of interest," etc. Then, the score corresponding to the group is calculated, and the users are assigned to the group in descending order of the score. However, the population distribution of users who use the Internet may differ from the actual population distribution. For example, it is not possible to secure a sufficient number of users for a group corresponding to "cosmetics for women in their 50s", which is in high demand. In some cases. In particular, companies that want to deliver advertisements want to deliver advertisements to as many users as possible, and for users who have attributes that can be presumed to be "50s,""women," and "interested in makeup." I want to deliver advertising content. Therefore, the conventional method of selecting only users who are likely to match "cosmetics for women in their 50s" cannot meet such a demand.
In contrast, in the present embodiment, in step S5, the corrected score S _ug using the correction value b _g, since selecting the group based on the corrected score T _ug, number of users U _g of each group However, each user is grouped so as to be close to the target value _{Ugt of the group.} In this case, when setting groups only by example score S _ug, excessively when the number of users U _g is large with respect to the target value U _gt groups, the group setting according to the correction score T _ug is performed Then, a user who is more suitable for the group is selected. On the other hand, when setting groups only by score S _ug, when a small number of users U _g is the target value U _gt group by group setting according to the correction score T _ug is performed, the group A user who is likely to meet the above conditions even a little will be selected as a user who belongs to the group.
As described above, in the present embodiment, _{it is possible to appropriately classify the users according to the target value Ugt} of each group. For example, in marketing activities such as distribution of advertising contents, it is appropriate according to the demand of the market. It becomes possible to select a user.

In the present embodiment, in the group setting unit 133, the ratio (U _g / U _{gt ) of the number of users U g} belonging to the group to the target value U _gt does not satisfy the equation (3) (out of the predetermined range). If so, the score _Sug is corrected.
When classifying users into a plurality of groups, it is most preferable that the threshold values P1 and P2 are set to P1 = P2 = 1.0, and the number of users U _g and the target value U _{gt completely match.} However, in order to accurately match the number of users U _g and the target value U _gt , it is necessary to set the index α of the correction coefficient to a small value and repeat the loop processing of steps S5 to S8 a very large number of times. Therefore, the processing time related to the classification processing becomes long, and the processing load also increases. On the other hand, as described above, the lower limit threshold value P1 and the upper limit threshold value P2 are set, _{and the score is corrected so that the ratio (Ug} / _Ugt ) is within the predetermined range, thereby shortening the processing time and reducing the processing time. , The processing load can be reduced.

Here, the threshold values P1 and P2 can be set individually for each of the groups.
The importance of the target value _{Ugt may differ for each group.} For example, to illustrate the distribution of advertising content, Some advertisers desire the delivery of advertising content to the user number U _g in ± 20% range with respect to the target value U _gt, ± with respect to the target value U _gt 5 % Some advertisers want the delivery of advertising content to the user the number of U _g of the range. In the present embodiment, by separately providing the threshold values P1 and P2 for each group, the demands of various advertisers can be suitably satisfied. That is, the users can be appropriately grouped according to the conditions set for each group.

Further, the threshold values P1 and P2 may be changed depending on the magnitude of the target value _Ugt.
That is, in the group target value U _gt large, the score of the previous number of users U _g is small with respect to the target value U _gt, and you are greatly different with respect to the target value U _gt, which is corrected for the group _{Many users with a small Sug} are more likely to be assigned to the modified group, which reduces the accuracy of the user's suitability for the group. Further, the group target value _{U gt} smaller, compared to the group target value _{U gt} large, the effect of per user for the ratio _(U g _/ U _gt) increases. Generally, in fields where demand is low in the market, the target value of the number of users in the group is set small. In such a field, since the number of users who are interested in the field is limited, it is necessary to select the user accurately. At this time, the threshold values P1 and P2 are set to values close to 1. As a result, users with attributes close to the conditions of each group can be appropriately assigned to the group.

_{In the present embodiment, when the ratio of the number of users U g} belonging to the temporary group and the target value _Ugt is out of the predetermined range (when it is determined as No in step S7), the score correction unit 133B is assigned to the temporary group. _{The correction value is sequentially corrected until the ratio of the number of users U g} to which the user belongs and the target value U _gt falls within a predetermined range in all the groups (until it is determined to be Yes in step S7). That is, in the present embodiment, by repeatedly executing the loop processing of steps S5 to S8, the number of users U _g and the target value U _gt become the same value or close values in all the groups. This dramatically improves the accuracy of user grouping.

In the score correction unit 133B of the present embodiment, when the ratio (U _g / U _gt ) is out of the predetermined range and the number of users U _g > the target value U _gt , the correction value b _g is increased, and conversely. When the number of users U _g <target value U _gt , the correction value b _g is reduced. Thus, when setting the temporary group based on the corrected score T _ug, it can be brought close to the number of users U _g target value U _gt group.

In this case, the score correction unit 133B, to correct the correction value _{b g} based on equation (4). That is, the correction value is increased or decreased according to the ratio between the number of users U _g and the target value U _gt. In this case, when the difference between the number of users U _g and the target value U _gt is large, the correction value b variation of _g (correction amount) is large, the correction score T _ug of change (correction score T the previously calculated _{(Difference from ug} ) also increases. Further, when the difference between the number of users U _g and the target value U _gt is small, the correction value b variation of _g (correction amount) is small, even smaller variation of the correction score T _ug.
That is, when the difference between the number of users U _g with respect to the target value U _gt large acts as the number of users U _g is rapidly converged to the target value U _gt, the number of users U _g with respect to the target value U _gt When the difference is small, it acts so that the _{number of users U g} approaches the target value U _{gt with high accuracy.} Therefore, it is possible to reduce the processing time related to the grouping (classification processing) of users and to improve the classification accuracy at the same time.

In the present embodiment, the correction value b _g which is set for each of a plurality of groups, calculates the correction score T _ug for each group for each user. That is, according to the magnitude of the ratio of the target value U _gt user number U _g, the correction value b _g for each group is set. Thus, it is possible to bring the number of users U _g belonging to each group to the target value U _gt.
For example the ratio in group A _(U g _/ U _gt) is within a predetermined range, the ratio in group B _(U g _/ U _gt) is in the case where a predetermined range, the correction value _{b g} for group A is maintained as it is ( No change in the correction score), only the correction value b _g for group B is modified (corrected scores are modified).
By correcting the correction score of group B, not only the number of users belonging to group B but also the number of users belonging to group A may change. In that case, the correction score for group A is corrected next. Will be done. By repeating the processes of steps S5 to S8 in this way, the number of users in all groups is finally optimized.

In this embodiment, the group setting unit 133, based on the corrected score T _ug, select the temporary group to contain users, the ratio of the number of users U _g and the target value U _gt belonging to each formal group within a predetermined range When becomes, the provisional group is registered as the group to which the classification target belongs.
In the present embodiment, a plurality of users are grouped into a plurality of groups so that the number of users _{U g} in each group approaches the target value U _gt. Therefore, even if the ratio ( _Ug / _Ugt ) is within a predetermined range in one group, the users belonging to the group may be changed when optimizing the other group. In this case, if _{the group ID of the group is registered in the user information when the ratio (Ug} / _Ugt ) is within a predetermined range in one group, the user belonging to the group is changed again when the user belongs to the group. It becomes necessary to update the user information. On the other hand, in the present embodiment, it is first determined whether or not the grouping is appropriate based on the number of users and the target value for each group, and then when it is determined that it is appropriate for all groups. , Associate each user with each group. As a result, unnecessary update processing as described above can be omitted, and the processing load can be reduced and the processing speed can be increased.

[Second Embodiment]
Next, the second embodiment according to the present invention will be described.
In the first embodiment described above, the score correction unit 133B, in step S8, in modifying the correction value b _g, and a fixed value index of the correction coefficient alpha. That is, in the first embodiment, since a fixed value correction coefficient, the index of the correction coefficient α is large, the change amount of the correction value b _g (correction amount) is increased, also the change value of the correction score T _ug larger Become. In this case, _{it is possible to quickly converge the number of users U g} to the target value U _gt , but conversely, there is a high possibility that the number of users will diverge without converging. On the other hand, if the index of the correction coefficient α is small, the correction value b _g variation is small, even smaller variation of the correction score T _ug. In this case, _{it is possible to accurately converge the number of users U g} to the target value U _gt , but conversely, the time required for processing becomes long.
On the other hand, the second embodiment differs from the first embodiment in that the index α of the correction coefficient does not become a fixed value.
In the following description, the same reference numerals will be given to the matters already described, and the description thereof will be omitted or simplified. Further, the configuration of the information processing system 1, the configuration of the server device 10, and the configuration of the terminal device 20 in the second embodiment are the same as those in the first embodiment, only the processing contents in the score correction unit 133B are different. Therefore, the processing performed by the score correction unit 133B will be mainly described below.

In a second embodiment, the score correction unit 133B, a difference between the number of users Ug and the target value U _gt, or depending on the ratio of the number of users U _g and the target value U _gt, changes the index of the correction coefficient alpha.
FIG. 5 is a flowchart showing the details of the process of step S8 (see FIG. 4; correction value correction process) in the second embodiment.
In the present embodiment, the score correction unit 133B first introduces a variable j indicating the number of corrections of the index α of the correction coefficient, and sets j = 1 as an initial value.
In the classification process shown in FIG. 4, the score correction unit 133B calculates the _{correction score Tug in the first step S5 with the correction value bg} = 1 as in the first embodiment.
Then, if No is determined in step S7, the score correction unit 133B executes the correction value correction process in step S8.
In this correction value correction process, in the present embodiment, the score correction unit 133B first determines whether or not the variable j is j = 1 (step S11).
If it is determined to be Yes in step S11 (when j = 1), the score correction unit 133B sets the index α of the correction coefficient to α = 1 (step S12).

On the other hand, when it is determined as No in step S11 (when j> 1), the score correction unit 133B is the user of each group (each temporary group) when the temporary group was selected for each user in the previous step S6. The ratio of the number U _{g (j-1)} to the target value U _{gt (U g (j-1)} / U _gt ) is calculated. Note that the last step S6, indicating the selection process of the temporary group based on the corrected score T _ug calculated using the correction value b _g set by modifying the index α of j-1 th correction factor.
Then, the score correction unit 133B, the ratio _{(U g (j-1)} / U gt) is, whether less than the first threshold Q1, or the ratio _{(U g (j-1)} / U gt) is It is determined whether or not it is larger than the second threshold value Q2 (step S13). The first threshold value Q1 is a value smaller than the lower limit threshold value P1, and the second threshold value Q2 is a value larger than the upper limit threshold value P2.

If Yes is determined in step S13, that is, if the ratio (U _{g (j-1)} / U _gt ) is smaller than the first threshold value Q1 or larger than the second threshold value Q2, the target predetermined range is still reached. Since it is far away, the score correction unit 133B resets the index α of the correction coefficient to nα (step S14). Here, n is a value greater than 1 (for example, 2). That is, the value of the exponent α of the correction coefficient used in the calculation (Equation (4) refer) correction value b _g in the previous step S5 is increased n times, the calculation of the correction value b _g at the next step S5 Let it be the index α of the correction coefficient used. In this embodiment, "n" is a fixed value.

On the other hand, when No is determined in step S13, that is, when the ratio (U _{g (j-1)} / U _gt ) is between the threshold values Q1 and Q2, the score correction unit is close to the target predetermined range. In 133B, the index α of the correction coefficient is reset to α / n (step S15). That is, the value of the exponent α of the correction coefficient used in the calculation of the correction value b _g (formula (4) refer) in the previous step S5 to 1 / n times, the calculation of the correction value b _g at the next step S5 Let it be the index α of the correction coefficient used.
After the above, the score correction unit 133B on the basis of the equation (4), calculates the correction value _{b g} (step S16), and adds "1" to the variable j (step S17).

When changing the exponent α of the correction coefficient, as described above, the correction score T _ug calculated by the score correction unit 133B, will be described with reference to specific examples.
FIG. 6 is a diagram showing an example of the transition _{of the ratio (Ug} / _{Ugt) in the present embodiment.}
In the example shown in FIG. 6, the correction score _Tug (= _{Sug ) calculated as the first time (b g} = 1) is small, and when a temporary group is selected for each user based on the _{correction score Tug, the ratio is calculated.} (U _gt / U _g0 ) becomes smaller than the first threshold value Q1. In this case, if No is determined in the first step S7, j = 1, so that the index α = 1 of the correction coefficient is set in step S13 of step S8, and the correction value b _g is b _g × (in step S16). It _{is corrected to U g0} / U _gt ), and the process returns to the process of step S5.

However, in this embodiment, even when calculating the correction score _{T ug} using the corrected correction value _{b g,} the ratio _(U g1 _/ U _gt) is smaller than the first threshold value Q1. Therefore, it is determined as Yes in step S13 of step S8, and the index α = 2 of the correction coefficient (assuming n = 2 in this example). As a result, in step S16, the correction value b _g is corrected to b _g × (U _g1 / _Ugt ) ^2, and the process returns to step S5. In this case, when the number of users U _g2 is calculated _{based on the correction value b g , U g2-} U _g1 > U _g1- U _g0 , and as shown in FIG. 6 (A), the ratio (U _g / U _gt ). Is very close to "1".
Further, in the present embodiment, even when calculating the correction score _{T ug} using the correction value _{b g,} the ratio _(U g2 _/ U _gt) is smaller than the first threshold value Q1. Therefore, it is determined as No in the third step S7, the index α = 4 of the correction coefficient is set in step S14, and the correction value b _g is corrected to b _g × (U _g2 / U _gt ) ⁴ in step S16. Will be done. In this case, U _g3- U _g2 > U _g2- U _g1 , and as shown in FIG. 6 (A), the ratio (U _g3 / U _gt ) further approaches "1".

As a result, although it is determined as No in step S7 of the loop processing, the ratio (U _g3 / _Ugt ) becomes a value between the first threshold value Q1 and the lower limit threshold value P1. Therefore, in the next step S8, it is determined as No in step S13, the index α of the correction coefficient is corrected to α / 2, that is, α = 2 in this example, and the correction value b _g is b _g. It is corrected to × (U _g3 / U _gt ) ^2.
As a result, the ratio (U _g3 / _Ugt ) becomes a value between the lower limit threshold value P1 and 1 (or the upper limit threshold value P2), and is determined as Yes in step S7 of the loop processing.

_{Further, in FIG. 6A} , the ratio (U) is based on the fifth correction score Tug calculated when the variable j becomes j = 4 (correction of the index α and the correction value of the fourth correction coefficient). _{This is an example in which g4} / _Ugt ) falls between the lower limit threshold value P1 and the upper limit threshold value P2. On the other hand, FIG. 6B shows a case where, for example, the ratio (U _g4 / _Ugt ) exceeds the range from the lower limit threshold value P1 to the upper limit threshold value P2 and becomes a value between the upper limit threshold value P2 and the second threshold value Q2. There is.
Even in such a case, the index α of the correction coefficient is further multiplied by 1 / n by determining No in the next step S13 (α = 1 in this example). As a result, as shown in FIG. 6B, the ratio (U _g5 / _Ugt ) becomes a value between the lower limit threshold value P1 and the upper limit threshold value P2, and the values converge.

[Action and effect of this embodiment]
In this embodiment, the score correction unit 133B compares the ratio of the number of users U _g and the target value U _gt belonging to the group (tentative Group), the value of the first threshold value Q1 and the second threshold value Q2, compared The value of the index α of the correction coefficient is increased or decreased according to the result. That is, in the case of (U _g / U _gt ) <Q1 or (U _g / U _gt )> Q2, that is, when the ratio is farther than a predetermined range, the index α of the correction coefficient is increased and Q1 <(U _g). / U _gt ) <P1 or P2 <(U _g / U _gt ) <Q2, that is, when the ratio is close to a predetermined range, the index α of the correction coefficient is reduced.
Thus, when the target value U _gt user number U _g is largely deviated, the rapidly using larger α to approach the target value U _gt number of users U _g, The target value U _gt When the number of users U _g and the number of users U g are close to each other, a small α is used so that the number of users U _g approaches the target value U _{gt with high accuracy.} Therefore, the server device 10 can quickly and accurately perform the classification process.

[Modification example]
The present invention is not limited to the above-described embodiment, but also includes the modifications shown below to the extent that the object of the present invention can be achieved.

(Modification example 1)
In the second embodiment, an example in which the ratio of the number of users U _g and the target value U _gt, based on a result of comparison between the first threshold value Q1 and the second threshold value Q2, increases or decreases the index of the correction coefficient α However, it is not limited to this.
For example, in addition to the comparison result between the _{ratio (U g} / U _gt ) and the threshold values Q1 and Q2, the index α of the correction coefficient may be increased or decreased based on the increasing _{/ decreasing tendency of the ratio (U g} / U _gt).
That, in the second _embodiment, to increase the index of the correction coefficient α in the case of _{(U g / U gt) <} Q1 or _{(U g / U gt)>} Q2. In this case, for example, the ratio ( _Ug / _Ugt ) may fluctuate from a value smaller than the first threshold value Q1 to a value exceeding the second threshold value Q2. In the second embodiment, in this case, ( _Ug / _Ugt )> Q2, so that the index α of the correction coefficient is further increased. Therefore, the next time a temporary group for the user is selected, ( U _g / U _gt ) <Q1, and the index α of the correction coefficient further increases, and the number of users U _g may _{not converge to U gt} and may diverge.
On the other hand, the score correction unit 133B determines, for example, _{the case where the value of the ratio (U g} / U _gt ) changes over 1, that is, the magnitude relationship between the _{number of users U g} and the target value U _gt. If the determination is made and the relationship is reversed and changed, the index α of the correction coefficient may be reduced (for example, corrected to α / n).

Further, α may be fixed for each range of the ratio (U _g / U _gt). For example, the third threshold value Q3 and the fourth threshold value Q4 where P1>Q1>Q3> Q4, the fifth threshold value Q5 and the sixth threshold value Q6 where P2 <Q2 <Q5 <Q6 are further set. Then, the score correction unit 133B _{sets α = 4 in the case of (U g} / U _gt ) <Q4 or (U _g / U _gt )> Q6, and Q4 ≦ (U _g / U _gt ) <Q3 or When Q5 <(U _g / U _gt ) ≤ Q6, set α = 3, and when Q3 ≤ (U _g / U _gt ) <Q1 or Q2 <(U _g / U _gt ) ≤ Q5, α = 2 is set, and α = 1 is set when _{Q1 ≦ (U g} / U _gt ) <P1 or P2 <(U _g / U _{gt) ≦ Q2.} That is, the _{closer the ratio (U g} / U _gt ) is to 1, the smaller α is set.
Further, the score correction unit 133B sets _{the absolute value of the difference between the ratio (U g} / U _gt ) and 1 as β (β = | 1- (U _g / U _gt ) |), and depending on the value of β. The index α of the correction coefficient may be calculated using a function that increases or decreases the index α of the correction coefficient (α = f (β), f is a non-decreasing function of β).
Even in these cases, _{the inconvenience of diverging the ratio (Ug} / _Ugt ) as described above can be suppressed, and a quick and highly accurate classification process can be performed.

(Modification 2)
In the second embodiment, the score correction unit 133B multiplies the index α by n when increasing the index α of the correction coefficient, and divides the index α by n when decreasing the index α of the correction coefficient. , The method of increasing / decreasing the index α of the correction coefficient is not limited to this. For example, the score correction unit 133B adds a predetermined value W1 to the index α (α = α + W1) when increasing the index α of the correction coefficient, and decreases the index α of the correction coefficient from the index α. W2 may be subtracted (α = α-W2).

(Modification example 3)
In the first embodiment, the lower limit threshold value P1 and the upper limit threshold value P2 are set to different values for each group, but for example, when the target values _Ugt in each group are close to each other, they may be common values.
Further, in the first embodiment, as the threshold values _{P1 and P2, an example in which the threshold values P1 and P2 of the group having a small target value Ugt} are set to a value close to 1 is shown, but the present invention is not limited to this. For example, the _{larger the target value Ugt} , the closer the threshold values P1 and P2 may be to 1. That is, in a _{group having a large target value U gt} , the number of users U _g is generally also large. In this case, if the threshold values P1 and P2 are far from 1, the number of users belonging to the group also fluctuates greatly. For example, when the lower limit threshold value is 0.8 and the upper limit threshold value is 1.2 for the target value “100 users”, the error range is up to ± 20 people. On the other hand, if the same threshold value is set for the target value "10,000 users", up to ± 2,000 people will be set as the error range. Therefore, when viewed as a ratio, it is the same 20%, but there will be a large difference in the actual number of users. On the other hand, as described above, the _{larger the target value Ugt} is, the closer the threshold values P1 and P2 are to 1, so that the above-mentioned inconvenience can be suppressed.

(Modification example 4)
In the first embodiment and the second embodiment, based on the ratio of the number of users U _g and the target value U _gt, an example has been shown to modify the index of the correction value b _g and the correction coefficient alpha, but not limited to .. That is, in the present invention, the process of correcting the score based on the information of the classification target corrects the score based on the condition of the classification target (ratio of the classification target belonging to the group and the target value) shown in the above embodiment. It is not limited to the one, and includes, for example, the one that corrects the score based on the number of classification objects belonging to the group itself.
As a specific example, for example, in such case the target value in each group is the same value or substantially the same value, based on the number of users U _g, by modifying the index of the correction value b _g and the correction coefficient α May be good. More specifically, the score correction unit 133B may correct the score based on the difference between the _{number of users U g} and the target value U _gt. In this case, in step S7, instead of the equation (3), _{it is determined whether or not P4 <Ugt} −U _g <P5, and if it is out of the range, the correction value _bg or the index of the correction coefficient is obtained. You may modify α.
The same applies with respect to the index of the correction coefficient alpha, the score correction unit 133B may modify the index of the correction coefficient alpha based on the difference between the number of users U _g and the target value U _gt.

(Modification 5)
In addition, an example of correcting the score based on the information of the classification target belonging to the group has been shown, but the score is not limited to this. For example, the score may be corrected based on the information of the classification target that does not belong to the group.
For example, when classifying users into four groups from "A" to "D", the correction value for the "A" group may be set based on the number of users belonging to "B" to "D". good.
Specifically, in the group other than the "A" group, depending on the number of groups that the ratio of the number of users U _g and the target value U _gt becomes within a predetermined range, sets the correction value of the "A" group You may. For example, in the score correction unit 133B, _{when the ratio of the number of users U gA} of the “A” group (provisional group) to the target value U _{g At} is out of the predetermined range, the ratio in each group from “B” to “D” ( the correction value when _U g _/ U _gt) is out of the predetermined range, the ratio in each group of the "D" from the "B" _(U g _/ U _gt) is larger than the correction value when it is within a predetermined range do.
In addition, when _{the number or ratio of groups whose ratio (Ug} / _Ugt ) is within the predetermined range exceeds the predetermined number, the score of each group is not corrected any more, and the group is set on the spot. It may be confirmed. For example, in 80% or more of all groups, when the ratio (U _g / U _gt ) is within the predetermined range, even if the ratio (U _g / U _gt ) is out of the predetermined range, some groups remain. Even if it is, the group may be confirmed there.
In this case, by changing the score excessively, it is possible to suppress a decrease in the accuracy of the user's suitability in other groups (that is, how well the user attributes match the conditions of the group).

(Modification 6)
In a first embodiment, the score correction unit 133B, as _b g = 1 the initial value of the correction value _{b g,} but the score _{S ug} is corrected (calculate the correction score _{T ug),} the first correction score _{T ug} You may skip the calculation of.
That is, first selects a temporary group to which the user belongs by only the score S _ug, when the number of users U _g is greatly different the target value U _gt, next time, the correction score T _ug by score correction unit 133B May be calculated.

(Modification 7)
In a first embodiment, the score correction unit 133B, the correction value b _g, based on the correction value b _g used in the previous score correction, an example of modifying the equation (4), but not limited to ..
For example, the score correction unit 133B sets the initial value of the correction value to b _g0, and calculates _{the correction value b gi} at the time of the i-th correction by b _gi = b _g0 × (U _g / U _gt ) ^α. Here, b _g0 is a fixed value.
In this case, based on the corrected corrected scores T _ug, for the group (tentative Group) classification users, number of users U _g of each group gradually changed, can be converged to the target value U _gt.

In this case, depending on the number of corrections of the correction value b _g (variable i), it may be increased or decreased the value of alpha.
For example, the exponent α may be modified to α / i. In this case, the larger the number of modifications, the smaller the value of α.
At this time, as shown in the second embodiment and the first modification, the correction value of the index α of the correction coefficient may be changed based on the ratio of the _{number of users Ug} and the target value _Ugt. The number of users U _g can be quickly and accurately converged to the _{target value U gt.}

Furthermore, the recording unit 12, the relationship between the correction value _{b g} for the ratio _(U g _/ U _gt), for example recorded beforehand as a table data, the score correction unit 133B, the correction for the ratio _(U g _/ U _gt) The value b _g may be read from the table data.

(Modification 8)
In the first embodiment, in step S6, the group selection unit 133C shows an example in which G groups are selected as temporary groups in descending order of the _{correction score Tug.} In this case, the temporary group selected for each user is recorded in, for example, a temporary storage area in the memory. On the other hand, the group registration unit 133D may record (register) the group selected by the group selection unit 133C in the group information to which the user information belongs. In this case, when No is determined in step S7 and the processes of steps S5 and S6 are executed again, the group registration unit 133D may update the group information to which the user information belongs. In this process, it is necessary to update the user information every time the loop process of steps S5 to S8 is repeated, but the update process (process of step S9) when it is determined to be Yes in step S7 may be omitted. can.

(Modification 9)
In the first embodiment, an example is shown in which the processes of steps S5 to S8 are repeated until it is determined to be Yes in step S7, but the process is not limited to this. _{If the correction score Tug} is calculated so that the number of users is within a predetermined range with respect to the target value of each group, only one correction may be required.
For example, based on the results of past grouping (classification processing), the target value of each group, the number of users belonging to each group when the users are grouped only by the score, the user's belonging condition for each group, and the correction value. Based on the above, a group prediction model is generated with the group information as the input value and the correction value as the output value. Then, when performing grouping, each group information is input to the group prediction model as an input value, and the output correction value of each group is used to correct the _{score Sug for each group of the user.} In this case, by updating the group prediction model by machine learning, it is possible to improve the accuracy of the correction value b _g for the group. _{Then, by using the correction value based on such a group prediction model, the correction score Tug} is calculated so that the number of users is within a predetermined range with respect to the target value of each group. For example, the calculation of is only required once, and the classification process can be performed more quickly.

(Modification example 10)
In a first embodiment, the score correction unit 133B, although an example of calculating the correction score by dividing the score S _ug correction value b _g, but is not limited thereto.
For example, if the number of users U _g is greater than the target value U _gt, it is less than 1, and if the number of users U _g is smaller than the target value U _gt, it is greater than 1, the score is _Sug . You may multiply the values.
Further, the score correction unit 133B may calculate the _{correction score Tug by adding or subtracting the score Sug} by a predetermined correction value. In this case, the larger the difference between the number of users and the target value, the larger the correction value is used. Then, when the number of users exceeds the target value, the score is subtracted by the correction value, and when the number of users is less than the target value, the score is added by the correction value.

(Modification 11)
In each of the first embodiment and the second embodiment, each modification, an example in which the classification target of the present invention is a user is shown, but the present invention is not limited thereto. For example, each agricultural product may be classified into a plurality of groups, with the agricultural product or the like as the classification target and the production area, harvest time, size, weight, color, etc. of the agricultural product as attribute information. In this case, for example, it can be classified into ordinary agricultural products that are in high demand in ordinary households, high-grade agricultural products that are in high demand by wealthy people, agricultural products for processing used in processing in factories and the like.

Further, in the above embodiment, the server device 10 classifies each user who operates the terminal device 20, and the server device 10 functions as the information processing device of the present invention. However, the terminal device 20 is classified according to the classification target. Can function as the information processing apparatus of the present invention.
For example, the terminal device 20 may function as an information processing device for classifying image information acquired by a user (for example, an image or video downloaded from the Internet, a photographic image or video taken by an imaging camera, etc.). can. In this case, each image may be classified using the download date / time, shooting date / time, file size, file type, shooting scene, etc. of the image or moving image as attribute information. At this time, the terminal device 20 causes the terminal control unit 25 to function as an attribute acquisition unit and a group setting unit by reading and executing the classification processing program recorded in the terminal recording unit 24. Thereby, by the information processing method substantially the same as that of the above embodiment, each image can be classified into the group set by the user according to the target value set by the user.

In addition, the specific structure and procedure for carrying out the present invention can be appropriately changed to other structures and the like as long as the object of the present invention can be achieved.

1 ... Information processing system, 10 ... Server device (information processing device), 12 ... Recording unit, 13 ... Control unit, 131 ... User information acquisition unit (attribute acquisition unit), 132 ... Group information acquisition unit, 133 ... Group setting unit , 133A ... Score calculation unit, 133B ... Score correction unit, 133C ... Group selection unit, 133D ... Group registration unit.

Claims (8)

Equipped with a group setting unit that divides the classification target into groups
The group setting unit corrects the score indicating the degree of relevance of the classification target to the group with the correction value set for each group, and selects a predetermined number of groups for each of the classification targets in descending order of the score. The correction value is increased or decreased so that the ratio of the number of the classification targets belonging to the provisional group to the target value regarding the number of the classification targets belonging to the provisional group is within a predetermined range. An information processing apparatus, characterized in that, when the ratio in each of the provisional groups is within the range of the predetermined range, the provisional group is set as the group to which the classification target belongs. In the information processing apparatus according to claim 1,
The group setting unit corrects the score by dividing the score by the correction value, increases the correction value when the number of the classification targets belonging to the group is larger than the target value, and increases the correction value when the number is smaller than the target value. An information processing device characterized by reducing the correction value. In the information processing apparatus according to claim 2,
The group setting section, the target value U _gt, the number of U _g of the classified belonging to the _group, the correction value b _g1 when the ratio reaches the predetermined _range, the index of the correction coefficient As α, the correction value _bg is corrected by the following equation (1).

An information processing device characterized by this. The information processing apparatus according to any one of claims 1 to 3.
The group setting unit is an information processing device characterized in that the score is corrected by a correction value set for each of a plurality of the groups. In the information processing apparatus according to claim 3,
The group setting unit is an information processing device that corrects the index α of the correction coefficient based on the information of the classification target. In the information processing apparatus according to claim 5,
The group setting section, the information processing apparatus characterized by modifying the index α of the correction coefficient based on the number U _g and a ratio of the classification target and the target value U _gt. It is an information processing method that classifies the classification target into one of a plurality of groups by a computer.
The computer includes a group setting unit.
The group setting unit corrects the score indicating the degree of relevance of the classification target to the group with the correction value set for each group, and selects a predetermined number of groups for each of the classification targets in descending order of the score. The correction value is increased or decreased so that the ratio of the number of the classification targets belonging to the provisional group to the target value for the number of the classification targets belonging to the group is within a predetermined range. An information processing method, characterized in that, when the ratio in each of the provisional groups is within the range of the predetermined range, the provisional group is set as the group to which the classification target belongs. An information processing program that is read and executed by a computer.
An information processing program characterized in that the computer functions as the information processing device according to any one of claims 1 to 6.

Images