JP7576059B2 - Information processing system, method and program - Google Patents

Description

This disclosure relates to technology for calculating scores for users.

Conventionally, a judgment device has been proposed that includes a user information acquisition unit that acquires behavioral information indicating the user's behavior, and a creditworthiness judgment unit that judges the user's creditworthiness regarding his or her future ability to repay loans based on the behavioral information (see Patent Document 1). Also, a borrower creditworthiness judgment program that allows a financial institution to automatically judge the creditworthiness of a borrower is known (see, for example, Patent Document 2).

JP 2021-174039 A JP 2021-140210 A

Conventionally, technology has been proposed to calculate a user score that represents a user's trustworthiness, etc., based on the user's behavioral history. It has also been proposed to calculate the user score while taking into account the risks associated with the user. However, the attribute data treated as the attributes of the user when calculating the user score is limited to data that mainly indicates the environment in which the user is placed and the user's spontaneous behavior, and there is room for improvement in improving the accuracy of the user score.

In view of the above-mentioned problems, the present disclosure aims to reflect changes in a user's behavior resulting from operations performed on that user in the estimation of the user score.

An example of the present disclosure is an information processing system including an effect estimation means for estimating, based on a group of attribute data relating to a user, the effect that a predetermined operation for encouraging the user to perform a predetermined action will have on whether the user will perform the action, a risk estimation means for estimating, based on the group of attribute data, a risk based on the likelihood that the user will not perform the action, an attribute extension means for extending the group of attribute data based on the estimated effect and risk, and a user score estimation means for estimating a user score to be set for the user based on the extended group of attribute data.

The present disclosure can be understood as an information processing device, a system, a method executed by a computer, or a program executed by a computer. The present disclosure can also be understood as such a program recorded on a recording medium readable by a computer or other device, machine, etc. Here, a recording medium readable by a computer, etc. refers to a recording medium that stores information such as data and programs through electrical, magnetic, optical, mechanical, or chemical action and can be read by a computer, etc.

According to the present disclosure, it is possible to reflect changes in a user's behavior resulting from operations performed on that user in the estimation of the user score.

1 is a schematic diagram showing a configuration of an information processing system according to a first embodiment. 1 is a diagram illustrating an outline of a functional configuration of an information processing device according to a first embodiment. 4 is a simplified diagram relating to a causality score estimation process according to the first embodiment; FIG. 4 is a simplified diagram relating to a process of estimating a risk index according to the first embodiment; 5 is a simplified diagram relating to a user score estimation process according to the first embodiment; FIG. 2 is a simplified diagram illustrating the concept of a decision tree of a machine learning model employed in the first embodiment. 1 is a flowchart showing a flow of machine learning processing according to the first embodiment. 11 is a flowchart showing the flow of a user score estimation process according to the first embodiment. 13 is a simplified diagram relating to a user score estimation process according to the second embodiment. 13 is a simplified diagram relating to a user score estimation process according to the second embodiment.

First Embodiment
Hereinafter, an embodiment of an information processing device, a method, and a program according to the present disclosure will be described with reference to the drawings. However, the embodiment described below is merely an example of an embodiment, and the information processing device, a method, and a program according to the present disclosure are not limited to the specific configuration described below. In carrying out the present disclosure, a specific configuration according to the embodiment may be appropriately adopted, and various improvements and modifications may be made.

In this embodiment, an embodiment will be described in which the technology according to the present disclosure is implemented for a user score management system that manages a user score 241 that indicates some measure related to a user (e.g., credit, etc.). However, the technology according to the present disclosure can be widely used as a technology for estimating a user score 241, and the application of the present disclosure is not limited to the examples shown in the embodiment.

<System Configuration>
1 is a schematic diagram showing the configuration of an information processing system according to this embodiment. In the system according to this embodiment, an information processing device 1 and one or more service providing systems 5 are connected to each other so as to be able to communicate with each other. A user is a user of a service provided by the service providing system 5, and receives the service by accessing the service providing system 5 from a user terminal.

The information processing device 1 is a computer including a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a storage device 14 such as an EEPROM (Electrically Erasable and Programmable Read Only Memory) or a HDD (Hard Disk Drive), a communication unit 15 such as a NIC (Network Interface Card), etc. However, the specific hardware configuration of the information processing device 1 can be omitted, replaced, or added as appropriate depending on the embodiment. Furthermore, the information processing device 1 is not limited to a device consisting of a single housing. The information processing device 1 may be realized by multiple devices using so-called cloud or distributed computing technology, etc.

The information processing device 1 manages a user score 241 for each user and provides the user score 241 to the service providing system 5. The service providing system 5 can customize services for the target user according to the user score 241 provided by the information processing device 1.

The service providing system 5 is a computer equipped with a CPU, ROM, RAM, a storage device, a communication unit, an input device, an output device, etc. (not shown). Furthermore, these systems and terminals are not limited to devices consisting of a single housing. These systems and terminals may be realized by multiple devices using so-called cloud or distributed computing technology, etc.

The services provided by the service providing system 5 are, for example, map information services, credit card/deferred payment services, electronic money payment services, online shopping services, online reservation services, operation center services, etc. Note that "deferred payment" is not limited to services such as Buy Now Pay Later (BNPL), but includes any purchase of products/services by deferred payment. The services provided by the service providing system 5 are not limited to the examples in this embodiment. The service providing system 5 notifies the information processing device 1 of the attribute data group 200 of the user acquired from the user when the service is provided. Here, the attribute data group 200 of the user includes the service usage history data of the user. The content of the service usage history data varies depending on the content of the service, and may include, for example, history data of the user's location information, payment history data of credit card usage amount/deferred payment usage amount, electronic money usage history data, transaction history data, reservation history data, operation history data for the user from the operation center, etc.

In the system according to this embodiment, the service providing system 5 is made up of an operation center management system 5a and a credit card/deferred payment management system 5b, which are connected to each other so that they can communicate with each other. An operation terminal (not shown) is installed in the operation center to perform operations according to instructions from the operation center management system 5a, and an operator operates the operation terminal to perform operations for users. The users are credit card/deferred payment users, and deposit the credit card usage amount/deferred payment usage amount via a financial institution or the like, and the payment history data for the credit card usage amount/deferred payment usage amount is notified to the operation center management system 5a via the credit card/deferred payment management system 5b.

The operation center management system 5a is a system for managing an operation center for collecting receivables by urging the payment of overdue credit card charges and/or deferred payment charges. Normally, the payment of credit card charges/deferred payment charges is made by debiting the user's account on the monthly withdrawal date, or by the user depositing the amount by a specified date, but there are cases where the payment of the credit card charges/deferred payment charges is not completed by the specified date due to reasons such as insufficient funds in the user's account or the user not depositing the amount by the specified date. For this reason, in the past, operations such as calling customers or sending messages from the operation center (call center) were performed to urge the user to pay the credit card charges/deferred payment charges and collect receivables.

In general, operations performed on users are effective in avoiding defaults (non-performance of debt), and the more operations performed, the higher the debt recovery rate. However, on the other hand, the more operations performed, the higher the costs of operations, such as labor costs, system usage fees, and system maintenance costs. In view of the above problems, the system disclosed herein employs technology to reduce costs related to operations without lowering the debt recovery rate. Note that in this embodiment, an example will be described in which the specified operation is mainly a phone call to the user, but the content of the specified operation is not limited, and may be various operations to encourage the user to take a specified action.

FIG. 2 is a diagram showing an outline of the functional configuration of the information processing device 1 according to this embodiment. The information processing device 1 functions as an information processing device including an effect estimation unit 21, a risk estimation unit 22, an attribute extension unit 23, a user score estimation unit 24, and a machine learning unit 25, by a program recorded in the storage device 14 being read into the RAM 13 and executed by the CPU 11, which controls each piece of hardware included in the information processing device 1. Note that in this embodiment and other embodiments described below, each function included in the information processing device 1 is executed by the CPU 11, which is a general-purpose processor, but some or all of these functions may be executed by one or more dedicated processors.

3 is a simplified diagram of a causal score estimation process according to an embodiment. The effect estimation unit 21 estimates the effect of a predetermined operation on a user to prompt the user to perform a predetermined action on whether or not the user performs the action. In this embodiment, the predetermined action is a payment of the overdue credit card usage amount/post-paid settlement usage amount. The specific payment method is not limited, and may be a transfer to a designated account or a payment at a designated counter. In this embodiment, the predetermined operation is a call to remind the user to pay the overdue credit card usage amount/post-paid settlement usage amount. The call to the user may be an automatic call using a recorded or mechanical voice, or a call in which an operator (human) talks to the user. In this embodiment, the effect estimation unit 21 estimates the effect of the operation using an effect estimation model 210 that outputs a causal score 211 indicating the effect of the operation on the user in response to an input of an attribute data group 200 including one or more user attributes related to the target user.

FIG. 4 is a simplified diagram of the estimation process of the risk index 221 according to the embodiment. The risk estimation unit 22 estimates the risk based on the probability that the user will not perform the predetermined action described above. There is no limit to the method of expressing the risk, and various indices may be adopted. For example, the risk can be expressed using the probability that the debt will not be collected and will become a default. In this embodiment, the risk estimation unit 22 estimates the risk based on the probability that the user will not perform the action using a risk estimation model 220 that outputs a risk index 221 indicating the probability that the debt will not be collected and will become a default for the user in response to an input of an attribute data group 200 including one or more user attributes related to the target user. However, the risk may be estimated according to, for example, a predetermined rule without using the risk estimation model 220. For example, the risk may be obtained by holding a value corresponding to each user attribute or combination of user attributes in advance and reading out this value. In addition, an index other than the probability of default may be adopted as an index indicating the risk. For example, the risk may be classified (ranked) and the class (rank) may be used as an index.

The attribute extension unit 23 extends the attribute data group 200 for the user. The attribute data group 200 for the user includes attribute data including usage history data acquired from the operation center management system 5a and the credit card/postpaid settlement management system 5b. At this time, the attribute extension unit 23 extends the attribute data group 200 for the user based on the estimated effect (in this embodiment, the causal score 211) and risk (in this embodiment, the risk index 221). In this embodiment, the attribute extension unit 23 determines the estimated causal score 211 and risk index 221 as they are as part of the attribute data group 200 for the user, and extends the attribute data group 200. However, the attribute extension unit 23 may perform some processing (normalization, ranking, labeling, etc.) on the estimated causal score 211 and risk index 221 to make them attribute data, or may use other types of scores calculated using the estimated causal score 211 and risk index 221 as attribute data. Additionally, other machine learning models may be used to calculate other types of scores.

Here, the attribute data group 200 expanded by the attribute expansion unit 23 may include demographic attributes, behavioral attributes, or psychographic attributes. Demographic attributes are, for example, the user's gender, family structure, age, etc., behavioral attributes may be based on service usage history data, such as whether or not a cash advance has been used, whether or not a revolving payment has been used, deposit and withdrawal history for a specific account, and commercial transaction history for some product including gambling or lottery (which may include online transaction history in an online marketplace, etc.), and psychographic attributes are, for example, preferences for gambling or lottery. However, the available user attributes are not limited to the examples in this embodiment. For example, "time required for operation (calls, etc.)" and "amount used by credit card/amount used for deferred payment" from an operation center, etc. may also be used as attributes.

In addition to the attribute data described above, the attribute data group 200 further includes a causal score 211 estimated by the effect estimation unit 21 using the effect estimation model 210, and a risk index 221 estimated by the risk estimation unit 22 using the risk estimation model 220.

Figure 5 is a simplified diagram of a user score estimation process according to an embodiment. The user score estimation unit 24 estimates a user score 241 to be set for a user based on an attribute data group 200. In this embodiment, the user score estimation unit 24 estimates a user score 241 to be set for the user by inputting the user's attribute data group 200 into a user score estimation model 240. Here, the output value of the user score estimation model 240 is, for example, a user score 241 normalized/standardized with 0 as the minimum value and 1 as the maximum value.

The machine learning unit 25 generates and/or updates a user score estimation model 240 used for user score estimation by the user score estimation unit 24, an effect estimation model 210 used for effect estimation by the effect estimation unit 21, and a risk estimation model 220 used for risk estimation by the risk estimation unit 22. The user score estimation model 240 is a machine learning model that outputs a user score 241 indicating some measure related to the user (e.g., trust, etc.) when one or more attribute data (attribute data group 200) related to the target user is input. The effect estimation model 210 is a machine learning model that outputs a causal score 211 indicating the effect of the operation on the target user when data of one or more user attributes related to the target user is input. The risk estimation model 220 for risk estimation is a machine learning model that outputs a risk index 221 indicating the degree of risk based on the probability that the user will not perform an action when data of one or more user attributes related to the target user is input.

When generating and/or updating the user score estimation model 240, the machine learning unit 25 creates the user score estimation model 240 for each user based on teacher data in which the attribute data group 200 including the demographic attributes of the user is defined as an input value and the user score 241 related to the user is defined as an output value. As described above, the attribute data group 200 input to the user score estimation model 240 includes the causal score 211 estimated by the effect estimation unit 21 and the risk index 221 estimated by the risk estimation unit 22, and is combined with the user score 241 of the corresponding user and input to the machine learning unit 25 as teacher data. The user score 241 set in the teacher data may be a user score determined based on a rule base, or may be a user score set manually (annotated). It may also be a user score corrected by an administrator or the like after being output in the past by the user score estimation model 240.

A machine learning model generation framework that can be adopted as the effect estimation model 210, the risk estimation model 220, the user score estimation model 240, etc. when implementing the technology disclosed herein is based on an ensemble learning algorithm, for example. For example, a machine learning framework (for example, LightGBM) based on a gradient boosting decision tree (GBDT) may be adopted for the framework. In other words, the framework may be a machine learning framework based on a decision tree model that transfers the error between the correct answer and the predicted value between the previous and next weak learners (weak classifiers). The predicted value here refers to the predicted value of the user score 241, for example. In addition to LightGBM, the framework may adopt boosting methods such as XGBoost and CatBoost. According to a framework using a decision tree, a machine learning model with relatively high performance can be generated with less effort in parameter adjustment compared to a framework using a neural network. However, the framework for generating a machine learning model that can be used to implement the technology disclosed herein is not limited to the example in this embodiment. For example, instead of a gradient boosting decision tree, another learning device such as a random forest may be used as a learning device, or a learning device that is not called a weak learning device such as a neural network may be used. In particular, when a learning device that is not called a weak learning device such as a neural network is used, ensemble learning does not need to be used.

Figure 6 is a simplified diagram showing the concept of a decision tree of a machine learning model adopted as the effect estimation model 210, the risk estimation model 220, the user score estimation model 240, etc. in this embodiment. When adopting a gradient boosting machine learning framework based on a decision tree algorithm, the branching conditions of each node of the decision tree are optimized. Specifically, in the gradient boosting machine learning framework based on a decision tree algorithm, a user score 241 is calculated for each user group having attributes indicated by two child nodes branched from one parent node, and the branching conditions of the parent node are optimized so that the difference between these user scores 241 becomes large (for example, so that the difference becomes maximum or becomes equal to or greater than a predetermined threshold), that is, so that the two child nodes branch neatly. For example, when the attribute indicated as the branching condition of the node is age, the age set as the branching threshold may be changed, or the branching condition may be changed to an attribute other than age. Also, for example, when the attribute indicated as the branching condition of a node in the user score estimation model 240 is the causal score 211, the amount of the causal score 211 set in the branching threshold may be changed, or the branching condition may be changed to an attribute other than the causal score 211. Also, for example, when the attribute indicated as the branching condition of a node is the risk index 221, the amount or class (rank) of the risk index 221 set in the branching threshold may be changed, or the branching condition may be changed to an attribute other than the risk index 221. In this way, by recursively optimizing the branching conditions of all nodes in the decision tree, it is possible to improve the estimation accuracy of the user score 241 based on the attribute data group 200.

In generating and/or updating the effect estimation model 210, the machine learning unit 25 creates teacher data (machine learning data) in which, for each user attribute, a score based on a statistic related to the execution rate of an action (debt recovery rate) by a user who has received an operation among a plurality of users having a predetermined attribute and a statistic related to the execution rate of an action by a user who has not received an operation among a plurality of users is defined as a causal score 211 indicating the effect of the operation on a user having the attribute, based on the data acquired from the operation center management system 5a and the credit card/postpaid settlement management system 5b. Then, the machine learning unit 25 creates the effect estimation model 210 based on the teacher data. In this embodiment, as an example, when the content of the operation is a call to a user, the causal score 211 based on the difference between each statistic is calculated by the formula "(debt recovery rate when the user makes a call) - (debt recovery rate when the user does not make a call)", and the calculated causal score 211 is combined with the attribute data of the corresponding user and input to the machine learning unit 25 as teacher data. In this embodiment, an example in which an average value is used as a statistic is described. However, as the statistics, for example, statistical indicators such as the mode or median may be used. Here, the statistics relating to the execution rate of the action may be based on the past debt collection rate of each user within a specified period (for example, a specified month). In addition, in this embodiment, the causal score 211 is calculated based on the difference between each statistical amount, but if the difference does not show a statistically significant difference, the causal score 211 may be set to zero or approximately zero. Here, an existing statistical method may be used to determine whether or not there is a significant difference. For example, the standard error and confidence interval for a set of the average collection rate of users for each month can be taken into consideration, and the statistical significance of the change in the collection rate depending on whether or not a call is made can be taken into consideration. In this way, it is possible to estimate (calculate) the causal score 211 taking into consideration the variation in the average collection rate of users in the same group.

When creating the teacher data, since a user who has been called once cannot become a user who has not been called thereafter, it is possible to obtain only one of the recovery rate when the user is called or the recovery rate when the user is not called for one user. For this reason, teacher data on the effect of a call is created for each user group having a common attribute. That is, the causal score 211 indicating the effect of a call on a user group consisting of users having a certain common attribute is obtained by, for example, dividing a plurality of users having the common attribute into a first sub-user group to be called and a second sub-user group to not be called, calculating the average value of the debt recovery rate from the first sub-user group to be called and the average value of the debt recovery rate from the second sub-user group to not be called, respectively, and calculating the difference between these average recovery rates based on the above formula. For example, if the average value of the debt recovery rate from the first sub-user group to be called is 80% and the average value of the debt recovery rate from the second sub-user group to not be called is 70%, the causal score 211 indicating the effect of a call on the user group is "10".

Furthermore, in generating or updating the effect estimation model 210, there are no limitations on the machine learning model generation framework that can be adopted, but a gradient boosting machine learning framework based on a decision tree algorithm may be adopted, as in the generation and/or update of the user score estimation model 240 described above.

When generating and/or updating the risk estimation model 220, the machine learning unit 25 creates teacher data that defines, for each user attribute, a statistical amount (in this embodiment, an average value. However, for example, a statistical index such as a mode or a median may be used) related to the default occurrence rate of multiple users having a predetermined attribute as a risk index 221 indicating the degree of risk of the user having the attribute, based on the data acquired from the operation center management system 5a and the credit card/postpaid settlement management system 5b. Then, the machine learning unit 25 creates the risk estimation model 220 based on the teacher data. The calculated risk index 221 is combined with the attribute data of the corresponding user and input to the machine learning unit 25 as teacher data. In addition, in generating or updating the risk estimation model 220, a machine learning model generation framework that can be adopted is not limited, but a machine learning framework of gradient boosting based on a decision tree algorithm may be adopted, as in the generation and/or update of the user score estimation model 240 described above.

<Processing flow>
Next, a flow of processing executed by the information processing system according to the present embodiment will be described. Note that the specific contents and processing order of the processing described below are an example for implementing the present disclosure. The specific contents and processing order may be appropriately selected according to the embodiment of the present disclosure.

Figure 7 is a flowchart showing the flow of the machine learning process according to this embodiment. The process shown in this flowchart is executed at a timing specified by the administrator.

In steps S101 and S102, the effect estimation model 210 is generated and/or updated. The machine learning unit 25 calculates a causal score 211 for each of a plurality of user attributes (attribute data group 200) based on the user's attribute data, operation history data, and payment history data of the credit card usage amount/deferred payment usage amount previously accumulated in the operation center management system 5a and/or the credit card/deferred payment management system 5b, and creates teacher data including a combination of the user attributes (attribute data group 200) and the causal score 211 (step S101). Here, the operation history data includes data that can be used to grasp whether or not an operation has been performed for each user, and the payment history data includes data that can be used to grasp whether or not the credit card usage amount/deferred payment usage amount of the user has been paid (whether or not there is a default) for each user. Then, the machine learning unit 25 inputs the created teacher data into the effect estimation model 210, and generates or updates the effect estimation model 210 used for effect estimation by the effect estimation unit 21 (step S102). Processing then proceeds to step S103.

In steps S103 and S104, the risk estimation model 220 used for risk estimation is generated and/or updated. The machine learning unit 25 calculates a risk index 221 for each of a plurality of user attributes (attribute data group 200) based on the user's attribute data, operation history data, and payment history data of the credit card usage amount/deferred payment usage amount previously accumulated in the operation center management system 5a and/or the credit card/deferred payment management system 5b, and creates teacher data including a combination of the user attributes (attribute data group 200) and the risk index 221 (step S103). Then, the machine learning unit 25 inputs the created teacher data into the risk estimation model 220, and generates or updates the risk estimation model 220 used for risk estimation by the risk estimation unit 22 (step S104). After that, the process proceeds to step S105.

In steps S105 and S106, the user score estimation model 240 is generated and/or updated. The machine learning unit 25 creates training data including a combination of attribute data group 200 for each user previously accumulated in the operation center management system 5a and/or credit card/postpaid settlement management system 5b and a user score 241 determined in advance for the corresponding user (step S105). Then, the machine learning unit 25 inputs the created training data into the user score estimation model 240, and creates or updates the user score estimation model 240 used for user score estimation by the user score estimation unit 24 (step S106). After that, the processing shown in this flowchart ends.

Figure 8 is a flowchart showing the flow of the user score estimation process according to this embodiment. The process shown in this flowchart is executed for each target user at a timing specified by the administrator.

In steps S201 and S202, a risk based on the effect of the operation and the probability that the user will not execute the action is estimated. The risk estimation unit 22 inputs data on one or more user attributes (attribute data group 200) related to the target user to the risk estimation model 220 generated and/or updated in step S104 for each of the multiple users, and obtains a risk index 221 corresponding to the user as an output from the risk estimation model 220 (step S201). In addition, the effect estimation unit 21 inputs data on one or more user attributes (attribute data group 200) related to the target user to the effect estimation model 210 generated and/or updated in step S102 for each of the multiple users, and obtains a causal score 211 corresponding to the user as an output from the effect estimation model 210 (step S202). Thereafter, the process proceeds to step S203.

In steps S203 and S204, the user score 241 is estimated and output. The attribute extension unit 23 extends the attribute data group 200 for the user based on the causal score 211 and risk index 221 estimated in steps S201 and S202 (step S203). In this embodiment, the causal score 211 and the risk index 221 are determined (added) as part of the attribute data group 200 (new user attributes) as they are, thereby extending the attribute data group 200. Therefore, the attribute extension unit 23 adds the causal score 211 and the risk index 221 estimated in steps S201 and S202 to the attribute data group 200 that is previously held by acquiring the attribute data group 200 for the target user from the operation center management system 5a or the credit card/post-payment settlement management system 5b, thereby extending the attribute data group 200 for the user. Then, the user score estimation unit 24 estimates a user score 241 based on the attribute data group 200 expanded in step S203, and outputs the user score 241 to the operation center management system 5a (step S204). In this embodiment, the user score estimation unit 24 inputs the attribute data group 200 including the causal score 211 and risk index 221 of the target user to the user score estimation model 240, and acquires the output value as the user score 241 set for the user. However, the method of estimating the user score 241 is not limited to the example in this embodiment. For example, the user score 241 may include a value calculated by inputting the attribute data group 200 into a predetermined function that is not a machine learning model. Thereafter, the processing shown in this flowchart ends.

The user score 241 set for each user is provided to a service provision system 5, such as an operation center management system 5a and a credit card/postpaid settlement management system 5b, and is used for customizing the services provided to the target user by the service provision system 5. For example, the operation center management system 5a manages operations for the target user according to the user score 241, and the operation terminal executes operations according to instructions output by the operation center management system 5a.

＜Effects＞
According to this embodiment, it is possible to reflect a behavioral change of a user caused by an operation on the user in the estimation of the user score 241. In addition, by using various user attribute data, it is possible to estimate (calculate) the user score 241 with high accuracy even when some attribute data does not exist for the target user.

Second Embodiment
In the embodiment described above, an example has been described in which the user score 241 of the target user is estimated based on the attribute data group 200 expanded based on the estimated causal score 211 and the risk index 221, but the estimated causal score 211 and the risk index 221 are not limited to being used only for expanding the attribute data group 200, and may be used in the following manner, for example. Note that a description of the configuration roughly similar to that of the first embodiment described above will be omitted.

9 is a simplified diagram of the estimation process of the user score according to the second embodiment. The user score estimation unit 24 selects and determines one user score estimation model 240 from a plurality of user score estimation models (first user score estimation model 240A, second user score estimation model 240B, third user score estimation model 240C, ... n-th user score estimation model 240N (n is an integer)) according to the causal score 211 and/or risk index 221, and estimates the user score 241 (user score 241A, user score 241B, user score 241C, ... user score 241N) based on the attribute data group 200. Here, n may be > 1, n may be 2, or n may be 3, and there is no limit to the number of user score estimation models 240 indicated by n. In addition, the input attribute data group 200 may be extended by the causal score and/or risk index 221, or may not be extended by the causal score and/or risk index 221. Furthermore, each user score estimation model 240 is trained (learned) using teacher data in which the causal score 211 and/or risk index 221 are common or similar. When the causal score 211 and/or risk index 221 for the user to be estimated indicates a causal score 211 and/or risk index 221 within a predetermined range, the user score 241 is estimated using the user score estimation model 240 that has been trained in advance according to the causal score 211 and/or risk index 221 within that range.

FIG. 10 is a simplified diagram of the estimation process of the user score according to the second embodiment. The user score estimation unit 24 may determine the user score 241D by adjusting, correcting, amending, or otherwise modifying the user score 241 obtained by inputting the attribute data group 200 into the user score estimation model 240 based on the causal score 211 and/or the risk index 221 related to the user to be estimated. Here, when the causal score 211 exceeds a predetermined threshold, the user score estimation unit 24 may upwardly correct the user score 241 output by the user score estimation model 240 to determine the user score 241D. On the other hand, when the risk index 221 exceeds a predetermined threshold, the user score estimation unit 24 may downwardly correct the user score 241 output by the user score estimation model 240 to determine the user score 241D. When correcting the user score 231 output by the user score estimation model 240, the coefficient to be applied to the output user score 241, the score to be added, etc. may be appropriately determined based on the difference between the causal score 211 and/or the risk index 221 and a predetermined threshold value, etc.

<Other variations>
In the above-described embodiment, an example was described in which the operation for the user is a call, but the type of operation for the user is not limited to a call. For example, a message transmission may be adopted as the type of operation for the user. Note that the means for transmitting the message is not limited here, and an email system, a short message service (SMS), or a message transmission and reception service of a social networking service (SNS) may be used.

In the above-described embodiment, the attribute data group 200 is expanded based on the effect estimated for one type of operation (call), but the attribute data group 200 may be expanded based on the effect estimated for each of a plurality of types of operations (e.g., calling and sending a message), and the user score 241 may be determined based on the effect. In this case, the effect estimation unit 21 estimates a first effect of a first operation (e.g., calling) for the user to prompt the user to perform a predetermined action on whether the user will perform the action, and a second effect of a second operation (e.g., sending a message) for the user to prompt the user to perform a predetermined action on whether the user will perform the action, and the attribute expansion unit 23 may expand the attribute data group 200 for the user based on each of the estimated first effect (first causal score), second effect (second causal score), and risk, and may determine the user score estimation model 240 and estimate the user score 241. At this time, the effect estimation unit 21 may determine the causal score 211 based on the first causal score and the second causal score. Specifically, the effect estimation unit 21 may calculate, for example, a weighted average of the first causal score and the second causal score as the causal score 211.

In this case, an effect estimation model 210 for estimating the effect of an operation is also generated and updated for each type of operation. For example, if the first operation is making a call and the second operation is sending a message, an effect estimation model 210 for making a call and an effect estimation model 210 for sending a message may be generated and updated.

In the above-described embodiment, the difference between the action execution rate of the sub-user group that received the operation and the action execution rate of the sub-user group that did not receive the operation is used as the causal score 211. However, the causal score 211 may be calculated by other methods. For example, when generating and/or updating the effect estimation model 210 for effect estimation, the machine learning unit 25 may create the effect estimation model 210 based on teacher data in which a score based on a statistic related to the action execution rate of a user who has a predetermined attribute and has performed a predetermined reaction to the operation among multiple users having the predetermined attribute, and a statistic related to the action execution rate of a user who has not performed the predetermined reaction among multiple users, is defined as the causal score 211 related to the user having the attribute. For example, in this variation, the causal score 211 is calculated by the formula "(debt collection rate when the user performs the predetermined reaction) - (debt collection rate when the user does not perform the predetermined reaction)".

Here, the predetermined reaction is, for example, the user's response to a call by dialing, or a conversation with an operator when the user calls back on a call, a reply to a message, marking a message as read, etc. Furthermore, the content of the reaction may be a positive response to a payment, or the presence or absence of a response to the due date of payment, etc. If the user's reaction is a voice reaction, it is also possible to determine the user's emotions, etc. based on the user's voice and determine whether the reaction was positive or not.

In addition, an example in which the attribute data group 200 is expanded based on the estimated causal score 211 and risk index 221 described above, and an example in which the user score estimation model 240 or the user score 241 is determined have been shown, but instead of the causal score 211 and the risk index 221, the attribute data group 200 may be expanded and the user score estimation model 240 or the user score 241 may be determined based on one other index (score) based on the causal score 211 and the risk index 221. In this case, the one other index (score) may be determined on a rule basis by the information processing device 1 based on each of the weighted causal score 211 and the risk index, and may be determined by inputting the attribute data group 200 that has not been expanded to any machine learning model, and each of the weighted causal score 211 and the risk index.

1. Information processing device

Claims (10)

an effect estimation means for estimating an effect of a predetermined operation for a user for encouraging the user to perform a predetermined action on whether or not the user will perform the action as a causal score based on a statistical amount relating to a rate of execution of the action by users among a plurality of users who have received the operation and a statistical amount relating to a rate of execution of the action by users among the plurality of users who have not received the operation , based on a group of attribute data related to the user;
a risk estimation means for estimating a risk based on a probability that the user will not execute the action based on the attribute data group;
an attribute extension means for extending the attribute data group based on the estimated effect and the estimated risk;
a user score estimation means for estimating a user score to be set for the user based on the expanded attribute data group;
An information processing system comprising: the effect estimation means estimates an effect of the operation using an effect estimation model that outputs the causal score indicating an effect of the operation on the user in response to an input of one or more attributes related to the user;
The information processing system according to claim 1 . the effect estimation model is created based on teacher data in which a score based on a statistic relating to an execution rate of the action by a user who has received the operation among a plurality of users having a predetermined attribute and a statistic relating to an execution rate of the action by a user who has not received the operation among the plurality of users is defined as the causal score indicating an effect of the operation on a user who has the attribute;
The information processing system according to claim 2 . the effect estimation means estimates a first effect that a first operation for a user for prompting the user to perform a predetermined action has on whether or not the user will perform the action, and a second effect that a second operation for prompting the user to perform the predetermined action has on whether or not the user will perform the action;
the attribute extension means extends the attribute data group based on the estimated first effect, the estimated second effect, and the estimated risk;
The information processing system according to claim 1 . the risk estimation means estimates the risk using a risk estimation model that outputs a risk index indicating a probability that a receivable from a user will not be collected and will become a default, in response to an input of one or more attributes related to the user;
The information processing system according to claim 1 . The user score estimation means estimates a user score to be set for the user by inputting the group of attribute data of the user into a user score estimation model.
The information processing system according to claim 1 . The user score estimation means estimates the user score using the user score estimation model generated using a machine learning framework based on a gradient boosting decision tree.
7. The information processing system according to claim 6. the user score estimation means estimates a user score to be set for the user by using the user score estimation model generated using teacher data having an attribute data group including a demographic attribute of the user as an input value and the user score related to the user as an output value;
The information processing system according to claim 7. The computer
an effect estimation step of estimating an effect of a predetermined operation for a user for encouraging the user to perform a predetermined action on whether or not the user will perform the action as a causal score based on a statistical amount relating to a rate of execution of the action by a user who has received the operation among a plurality of users and a statistical amount relating to a rate of execution of the action by a user who has not received the operation among the plurality of users , based on a group of attribute data related to the user;
a risk estimation step of estimating a risk based on a probability that the user will not execute the action based on the attribute data group;
an attribute expansion step of expanding the attribute data group based on the estimated effect and the estimated risk;
a user score estimating step of estimating a user score to be set for the user based on the expanded attribute data group;
How to do it. Computer,
an effect estimation means for estimating an effect of a predetermined operation for a user for encouraging the user to perform a predetermined action on whether or not the user will perform the action as a causal score based on a statistical amount relating to a rate of execution of the action by users among a plurality of users who have received the operation and a statistical amount relating to a rate of execution of the action by users among the plurality of users who have not received the operation , based on a group of attribute data related to the user;
a risk estimation means for estimating a risk based on a probability that the user will not execute the action based on the attribute data group;
an attribute extension means for extending the attribute data group based on the estimated effect and the estimated risk;
a user score estimation means for estimating a user score to be set for the user based on the expanded attribute data group;
A program that functions as a

Images