Information processing device, information processing method, and program

Abstract

The purpose is to improve the efficiency of collection work for debtors. [Solution] The information processing device 1 of this embodiment is a machine learning model trained by machine learning using multiple sets of debtor attributes and effectiveness data indicating the effectiveness of reminders to encourage debtors to make payments as training data, and has a memory unit 12 that stores a classification model that outputs effectiveness data when debtor attributes are input, an effectiveness identification unit 131 that identifies the effectiveness of reminders for each of multiple target debtors based on the effectiveness data output by the classification model to which the attributes of the target debtors to be classified have been input, and a reminder target determination unit 132 that determines, from among the multiple target debtors, some target debtors with relatively high reminder effectiveness as reminder targets.

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] There is known a technique for improving the efficiency of debt collection operations (see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2008-9628 Summary of the Invention [Problem to be solved by the invention]

[0004] Among debtors, there are some for whom reminders are effective, and the probability of them making a payment increases when they are sent. On the other hand, there are also some for whom reminders are ineffective, such as those who make payments even without reminders or those who do not make payments even when reminders are sent. Financial institutions and other entities that have provided loans to debtors have limited resources that they can use to remind debtors. For this reason, it is important to send reminders efficiently, but there is a problem in that sending reminders to those for whom reminders are ineffective is inefficient.

[0005] The present invention has been made in consideration of these points, and aims to improve the efficiency of collection work for debtors. [Means for solving the problem]

[0006] The information processing device of the first aspect of the present invention is a machine learning model trained by machine learning using multiple sets of debtor attributes and effectiveness data indicating the effectiveness of reminders to encourage the debtor to make payment as training data, and has: a memory unit that stores a classification model that outputs the effectiveness data when the debtor attributes are input; an effectiveness determination unit that determines the reminder effectiveness for each of multiple target debtors based on the effectiveness data output by the classification model to which the attributes of the target debtors to be classified are input; and a reminder target determination unit that determines, among the multiple target debtors, some of the target debtors with relatively high reminder effectiveness as reminder targets.

[0007] The classification model may be configured to output, in association with each of the plurality of target debtors, first effectiveness data indicating that the probability of receiving the payment without a reminder is equal to or greater than a first threshold, second effectiveness data indicating that the probability of receiving the payment after a reminder is less than a second threshold that is smaller than the first threshold, or third effectiveness data indicating that the probability of receiving the payment without a reminder is less than the first threshold and that the probability of receiving the payment after a reminder is equal to or greater than the second threshold.

[0008] The dunning target determination unit may determine the target debtor corresponding to the third validity data as the dunning target.

[0009] The memory unit may store a dunning capability value indicating the ability of a dunning officer to carry out dunning work, and the dunning target determination unit may determine, among the target debtors corresponding to the second effectiveness data, those whose debt balance is equal to or greater than a threshold value as the dunning targets for whom the dunning officer will carry out the dunning work, whose dunning capability value satisfies a predetermined condition.

[0010] The dunning target determination unit may determine as the dunning targets some of the target debtors who have relatively high dunning scores calculated based on the dunning effectiveness and the debt balance of the debtor, among the multiple target debtors.

[0011] The memory unit may further store a payment prediction model that outputs the probability of payment for each dunning means when the debtor's attributes or the effectiveness data are input, and the information processing device may further have a dunning means determination unit that determines the dunning means for each of the dunning targets to be executed, the dunning means for which the payment probability output by the payment prediction model when the debtor's attributes or the output effectiveness data are input.

[0012] The dunning means determination unit may assign the total number of times the dunning can be carried out to the number of times the execution dunning means is executed for each of the dunning targets so that the value obtained by multiplying the probability of payment corresponding to the execution dunning means for each of the dunning targets by the debt balance of the target debtor and the number of times the execution dunning means is executed is maximized when added together for the number of dunning targets.

[0013] The dunning means determination unit may determine that the automatic call will be the execution dunning means if the difference between the probability of payment being made by a manual call by a dunning officer and the probability of payment being made by an automatic call, as output by the payment prediction model, is less than a threshold value.

[0014] The memory unit may further store a response probability prediction model that outputs the debtor's response probability to the reminder based on the timing of the reminder when the debtor's attributes are input, and the information processing device may further have a reminder timing determination unit that determines the timing for reminding each of the reminders to be the timing when the response probability output by the response probability prediction model to which the attributes of the reminder recipients have been input is relatively high.

[0015] The memory unit may store the availability status of the dunning staff by date and time, and the information processing device may further have a schedule generation unit that generates a dunning schedule by assigning dunning tasks for the date and time corresponding to the dunning timing to the dunning staff who are available to work on that date and time.

[0016] The schedule generation unit may generate the dunning schedule by assigning the dunning tasks to the dunning personnel in order of the dunning tasks for the dunning target person whose period between the dunning date made by the dunning personnel and the deposit execution date made by the dunning target person is shortest.

[0017] The memory unit may further store a response probability prediction model that outputs the response probability of the debtor to the dunning for each dunning means when the debtor's attributes are input, and the information processing device may further have a dunning means determination unit that determines the dunning means for each of the dunning targets, with the response probability output by the response probability prediction model when the attributes of the dunning targets are input, as the dunning means to be executed for each of the dunning targets.

[0018] An information processing device according to a second aspect of the present invention comprises a memory unit that stores an effectiveness table in which debtor attributes are associated with effectiveness data indicating the effectiveness of reminders to encourage the debtor to make payment; an effectiveness determination unit that determines the effectiveness of reminders for each of a plurality of target debtors based on the effectiveness data associated in the effectiveness table with the attributes of the debtor to be classified that have a similarity equal to or greater than a predetermined threshold; and a reminder target determination unit that determines, among the plurality of target debtors, some of the target debtors whose reminder effectiveness is relatively high as reminder targets.

[0019] An information processing method according to a third aspect of the present invention is a machine learning model trained by machine learning using as training data multiple sets of debtor attributes and effectiveness data indicating the effectiveness of reminders to encourage the debtors to make payments, and includes an effectiveness determination step executed by a computer having a memory unit that stores a classification model that outputs the effectiveness data when the debtor attributes are input, the step determining the effectiveness of reminders for each of multiple target debtors based on the effectiveness data output by the classification model to which the attributes of the target debtors to be classified have been input, and a reminder target determination step determining, as reminder targets, some of the target debtors among the multiple target debtors whose reminder effectiveness is relatively high.

[0020] A program relating to a fourth aspect of the present invention is a machine learning model that has been machine-learned using multiple sets of debtor attributes and effectiveness data indicating the effectiveness of reminders to encourage the debtor to make payment as training data, and causes a processor of an information processing device having a memory unit that stores a classification model that outputs the effectiveness data when the debtor attributes are input to function as an effectiveness determination unit that determines the effectiveness of reminders for each of multiple target debtors based on the effectiveness data output by the classification model to which the attributes of the target debtors to be classified are input, and a reminder target determination unit that determines, among the multiple target debtors, some of the target debtors with relatively high reminder effectiveness as reminder targets. [Effects of the Invention]

[0021] According to the present invention, it is possible to improve the efficiency of the work of urging debtors. [Brief explanation of the drawings]

[0022] [Figure 1] FIG. 2 is a diagram illustrating an outline of the operation of the information processing system S. [Figure 2] 10 is a flowchart showing the flow of processing from debtor classification to dunning schedule generation. [Figure 3] 1 is a diagram illustrating an example of a configuration of an information processing device 1. FIG. [Figure 4] FIG. 10 is a diagram illustrating an example of a reminder capability value table. [Figure 5] FIG. 10 is a diagram illustrating an example of a reminder classification table. [Figure 6] FIG. 10 is a diagram illustrating an example of a reminder means table. [Figure 7] FIG. 10 is a diagram illustrating an example of a reminder execution count table. [Figure 8] FIG. 10 is a diagram illustrating an example of a reminder timing table. [Figure 9] 10 is a flowchart showing a process flow relating to generation of a reminder schedule. [Figure 10]FIG. 10 is a diagram illustrating an example of a reminder schedule table. [Figure 11] FIG. 10 is a diagram illustrating an example of an effectiveness table. DETAILED DESCRIPTION OF THE INVENTION

[0023] [Outline of Information Processing System S] An overview of an information processing system S according to this embodiment will be described using Figures 1 and 2. Figure 1 is a diagram showing an overview of the operation of the information processing system S. The information processing system S includes an information processing device 1 and an information terminal 2. The information processing system S may also include other devices such as a server and a terminal.

[0024] Among debtors, there are delinquent debtors who do not make a payment by the payment date specified by the financial institution or other institution that provided the loan to the debtor. The number of such delinquent debtors is enormous, but financial institutions have had the problem of incurring significant costs and labor when it comes to urging payment even from delinquent debtors for whom dunning efforts are ineffective. In particular, when dunning personnel directly call debtors to urge them to make a payment, it is necessary to make effective use of the limited number of dunning personnel. Therefore, the information processing device 1 determines a portion of debtors for whom dunning efforts are relatively effective as dunning targets. This narrows down the number of debtors to be dunned, thereby improving the efficiency of dunning operations.

[0025] The information processing device 1 is a computer such as a server that determines who is the subject of a dunning request. The information terminal 2 is an information terminal used by a dunning representative or their manager. The information terminal 2 may be a stationary terminal such as a desktop personal computer, or a portable terminal such as a smartphone or tablet personal computer. The information processing device 1 and the information terminal 2 are connected to each other via a communication network such as the Internet.

[0026] An overview of the processing executed by the information processing device 1 will be described with reference to Fig. 1. The information processing device 1 stores a classification model that has been machine-learned using multiple sets of debtor attributes and effectiveness data indicating the effectiveness of reminders to encourage debtors to make payments as training data. When the attributes of a target debtor SA to be classified are input, the classification model outputs effectiveness data for the target debtor SA.

[0027] Based on the effectiveness data output by the classification model, the information processing device 1 classifies multiple target debtors SA into three groups: a no-reminder group that tends to make payments even without reminders, a valid reminder group that tends to make payments when reminders are sent, and an invalid reminder group that tends not to make payments even when reminders are sent.The information processing device 1 then transmits data indicating the valid reminder group as dunning target persons T to the information terminal 2 used by the dunning officer P.

[0028] The dunning officer P checks the dunning target T on the information terminal 2 and urges the dunning target T to make payment by telephone, email, etc. The number of dunning target T is small compared to the total number of delinquent debtors, and the dunning target T is effective for dunning, so the dunning officer P can carry out the dunning work efficiently.

[0029] Note that the dunning officer P will give priority to dunning the dunning target T who belongs to the valid dunning group, but for example, as soon as the dunning is completed or in parallel with the dunning, he may also make dunning requests to the target debtors SA who belong to the no-dunning group and the target debtors SA who belong to the invalid dunning group. In other words, it is not excluded that the dunning officer P may make dunning requests to target debtors SA other than the dunning target T. The dunning officer P may also make dunning requests to the guarantors of the dunning target T.

[0030] In order for the dunning person P to make a dunning request, it is desirable to have a dunning schedule that indicates which dunning person P will make a dunning request to which dunning target T, using what dunning means (telephone, email, etc.), how many times, and when. Therefore, the information processing device 1 may generate a dunning schedule. Below, a series of processes from classifying the target debtor SA to generating a dunning schedule will be described.

[0031] FIG. 2 is a flowchart showing the process flow from debtor classification to dunning schedule generation. First, as described above, the information processing device 1 classifies the target debtors SA into three groups: a dunning unnecessary group, a dunning effective group, and a dunning ineffective group, and determines the dunning target T from the dunning effective group (S1). Next, the information processing device 1 determines for each dunning target T, from among multiple dunning methods, a dunning method that has a relatively high probability of the dunning target T making a deposit (S2). Next, the information processing device 1 determines the number of times to execute the dunning method determined for each dunning target T in S2 so as to maximize the total amount recovered through dunning (S3). For example, the information processing device 1 determines the number of times to execute the dunning method for each dunning target T so as to execute the dunning method more frequently for dunning target T with a larger debt balance.

[0032] Next, the information processing device 1 determines, for each dunning target T, a dunning timing at which the probability that the dunning target T will respond to a dunning request by telephone, email, or the like is relatively high (S4). For each dunning target T, the information processing device 1 determines, for example, whether the dunning request should be sent to the dunning target T's landline or mobile phone, and at what time of day on a weekday or a holiday, at which the probability of a response is high. Note that the processing of S4 may be executed before S2 or S3, as long as it is executed after S1.

[0033] Finally, the information processing device 1 generates a dunning schedule indicating that the dunning person P will carry out the dunning means determined in S2 to the dunning target T determined in S1 for the number of times determined in S3 at the dunning timing determined in S4 (S5). The dunning person P carries out the dunning by referring to the generated dunning schedule.

[0034] [Configuration of information processing device 1] The following describes the configuration of the information processing device 1. Fig. 3 is a diagram showing an example of the configuration of the information processing device 1. The information processing device 1 includes a communication unit 11, a storage unit 12, and a control unit 13. The control unit 13 includes an effectiveness identification unit 131, a dunning target determination unit 132, a dunning means determination unit 133, a dunning timing determination unit 134, and a schedule generation unit 135.

[0035] The communication unit 11 is a communication interface for communicating with the information terminal 2 used by the dunning person P via a communication network such as the Internet. The communication unit 11 transmits the dunning classification table input from the dunning target determination unit 132 to the information terminal 2. The communication unit 11 also transmits the dunning schedule table input from the schedule generation unit 135 to the information terminal 2.

[0036] The storage unit 12 is a storage medium including a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The storage unit 12 stores a program executed by the control unit 13. The storage unit 12 stores an information processing program that causes the control unit 13 to function as, for example, an effectiveness specification unit 131, a reminder target determination unit 132, a reminder means determination unit 133, a reminder timing determination unit 134, and a schedule generation unit 135.

[0037] The memory unit 12 stores deposit history data. The deposit history data is data showing the past deposit history of a debtor and is the subject of learning by a machine learning model. The deposit history data associates information for identifying the debtor, the debtor's attributes, the deposit deadline, the date on which the debtor made a deposit, the deposit amount, and, if a dunning officer P made a dunning request, information for identifying the dunning officer P, the dunning method, and the date and time of the dunning request. The debtor's attributes include, for example, the debtor's employment status (part-time, temporary employee, company employee, etc.), marital status (single, married), residential status (living alone, living with parents, renting, owning a home), family composition, age, occupation, place of employment, work history, educational background, annual income, gender, personality, or financial institution usage history.

[0038] The memory unit 12 stores debtor list data. The debtor list data is data showing a debtor list indicating people who currently have debts. Debtors included in the debtor list are target debtors SA, who are the targets of classification using the classification model described below. The debtor data associates information for identifying the debtor with the debtor's attributes, the current debt balance, and the payment deadline.

[0039] The memory unit 12 stores a classification model for classifying multiple target debtors SA according to the effectiveness of dunning. The classification model is a machine learning model that is trained by machine learning using multiple sets of debtor attributes and effectiveness data indicating the effectiveness of dunning prompts to debtors as training data. The classification model is trained by machine learning, for example, by referencing payment history data and using debtor attributes, information indicating whether the debtor has made a payment, and, if a payment has been made, information indicating whether the payment was made before or after a dunning prompt from the debtor P. When debtor attributes are input, the classification model outputs effectiveness data. When debtor attributes are input, the classification model outputs, for each of multiple debtors, a pre-dunning payment probability, which is the probability that the payment will be made even without dunning, and a post-dunning payment probability, which is the probability that the payment will be made after dunning, as effectiveness data.

[0040] The memory unit 12 stores a payment prediction model for predicting the probability of receiving payment by each dunning method. The payment prediction model is machine-learned, for example, by referencing payment history data and using the debtor's attributes and information indicating whether or not a payment will be received when a dunning request is made using a predetermined dunning method. The payment prediction model may also be machine-learned using the effectiveness data output by the classification model and information indicating whether or not a payment will be received when a dunning request is made using a predetermined dunning method. Examples of dunning methods include manual calls by a dunning officer P, calls using a predictive dialing (PD) system or an automated call using an interactive voice response (IVR) system, email, short message service (SMS), or written letters or postcards. When the debtor's attributes or effectiveness data are input, the payment prediction model outputs the probability of receiving payment by each dunning method.

[0041] The storage unit 12 stores a response probability prediction model for predicting the probability of a debtor responding to a dunning request for each timing of the dunning request. The response probability prediction model is machine-learned, for example, by referencing the deposit history data and using the debtor's attributes, the dunning date and time, and information indicating whether or not the dunning target T will respond when the dunning request is sent on that dunning date and time. When the debtor's attributes are input, the response probability prediction model outputs the probability of a debtor responding to a dunning request for each timing of the dunning request.

[0042] The response probability prediction model may be machine-learned using information indicating whether the reminder was sent via a landline or mobile phone of the reminder recipient T. In this case, when the attributes of the debtor are input, the response probability prediction model outputs the probability of the debtor responding to a reminder via a landline or mobile phone reminder.

[0043] The memory unit 12 stores a dunning ability value table that indicates dunning ability values ​​that indicate the ability of dunning staff P to perform dunning work. FIG. 4 is a diagram showing an example of a dunning ability value table. In the dunning ability value table, dunning staff IDs are associated with dunning ability values. The dunning staff ID is information for identifying the dunning staff P. The dunning ability value is determined based on the number of years of dunning experience of the dunning staff P, the cumulative number of dunning attempts, the success rate of dunning, and the supervisor's evaluation of the dunning work, etc. The dunning ability value may be classified according to the skill level of the dunning staff P (expert level, intermediate level, beginner level) as shown in FIG. 4, or may be a numerical value such as a score or deviation value.

[0044] The storage unit 12 stores an availability status table that indicates the availability status of the dunning person P by date and time. The availability status table is generated, for example, by registering in the information processing device 1, at a timing such as the end of the month, the date and time when the dunning person P is available to perform dunning work in the following month.

[0045] The control unit 13 is, for example, a CPU (Central Processing Unit). The control unit 13 executes an information processing program stored in the storage unit 12, thereby functioning as an effectiveness identification unit 131, a dunning target determination unit 132, a dunning means determination unit 133, a dunning timing determination unit 134, and a schedule generation unit 135.

[0046] The effectiveness identification unit 131 identifies the dunning effectiveness for each of multiple target debtors SA based on the effectiveness data output by the classification model to which the attributes of the multiple target debtors SA to be classified are input. For example, when the attributes of the target debtors SA included in the debtor list indicated by the debtor list data stored in the memory unit 12 are input, the classification model outputs the pre-dunning payment probability and the post-dunning payment probability as effectiveness data, associated with each of the multiple target debtors SA. In this case, the effectiveness identification unit 131 identifies the dunning effectiveness for each of the multiple target debtors SA so that the dunning effectiveness of the target debtor SA having a larger value obtained by subtracting the output pre-dunning payment probability from the output post-dunning payment probability is higher.

[0047] When attributes of a target debtor SA included in the debtor list indicated by the debtor list data are input, the classification model may output first effectiveness data, second effectiveness data, or third effectiveness data in association with each of the multiple target debtors SA. The first effectiveness data is data indicating that the probability of making a payment before dunning is equal to or greater than a first threshold (e.g., 80%). The second effectiveness data is data indicating that the probability of making a payment after dunning is less than a second threshold (e.g., 60%) that is smaller than the first threshold. The third effectiveness data is data indicating that the probability of making a payment before dunning is less than the first threshold and the probability of making a payment after dunning is equal to or greater than the second threshold.

[0048] In this case, the validity identification unit 131 classifies the target debtor SA corresponding to the first validity data into a dunning-unnecessary group, the target debtor SA corresponding to the second validity data into a dunning-ineffective group, and the target debtor SA corresponding to the third validity data into a dunning-effective group. The validity identification unit 131 identifies the dunning validity for each of the multiple target debtors SA so that the dunning validity of the target debtor SA belonging to the dunning-effective group is higher than the dunning validity of the target debtor SA belonging to the dunning-unnecessary group and the dunning validity of the target debtor SA belonging to the dunning-ineffective group.

[0049] The dunning target determination unit 132 determines, from among multiple target debtors SA, some target debtors SA with relatively high dunning effectiveness as dunning target T. For example, from among multiple target debtors SA, the dunning target determination unit 132 determines, as dunning target T, the target debtor SA that corresponds to the third effectiveness data. The dunning target determination unit 132 may generate a dunning classification table that indicates the dunning classification group to which the target debtor SA belongs and the dunning target T.

[0050] Figure 5 is a diagram showing an example of a dunning classification table. In the dunning classification table, dunning classification groups, target debtor IDs, debt balances, debt target flags, and dunning target IDs are associated with each other. The target debtor ID is information for identifying the target debtor SA that is the subject of the classification. The dunning target flag is a flag for identifying the target debtor SA that is the subject of dunning, and in Figure 5, all target debtors SA that belong to the valid dunning group are flagged with a "○" indicating that they are the subject of dunning. The reason why the target debtor SA of SA009 and the target debtor SA of SA011 that belong to the invalid dunning group are also flagged with a "○" will be described later. The dunning target ID is information for identifying the dunning target T that is the subject of dunning.

[0051] The dunning target determination unit 132 may transmit the generated dunning classification table to the information terminal 2 used by the dunning person P. This allows the dunning person P to give priority to dunning the dunning target T, thereby improving the efficiency of dunning work for debtors. The dunning target determination unit 132 may store the generated dunning classification table in the storage unit 12. This allows the schedule generation unit 135 to generate a dunning schedule by referring to the dunning classification table, as will be described in detail later.

[0052] Incidentally, among the target debtors SA belonging to the dunning effective group corresponding to the third effectiveness data, there may be some whose probability of making a payment after the dunning is lower than the probability of making a payment before the dunning. If such persons whose probability of making a payment does not increase even after dunning are excluded from the dunning target persons T, the dunning officer P can make more efficient dunning efforts. Therefore, the dunning target determination unit 132 may determine as dunning target persons T those target debtors SA belonging to the dunning effective group whose probability of making a payment after the dunning is higher than the probability of making a payment before the dunning. This allows the dunning officer P to narrow down the dunning efforts to debtors for whom dunning is extremely effective.

[0053] Although the target debtor SA corresponding to the first validity data is highly likely to make a payment even without a reminder, there is a possibility that some of them will not make a payment. Therefore, the reminder target determination unit 132 may determine, among the target debtors SA corresponding to the first validity data, those who do not make a payment within a predetermined period (for example, one week) from the payment deadline as the reminder target T. This makes it possible to maximize the amount of debt collected.

[0054] In order to maximize the amount of debt recovered through the dunning service, the dunning target determination unit 132 may determine the dunning target T based on the debt balance of the target debtor SA in addition to the dunning effectiveness. For example, the dunning target determination unit 132 determines, as the dunning target T, some of the target debtors SA among multiple target debtors SA whose dunning scores calculated based on the dunning effectiveness and the debt balance of the debtor are relatively high. As an example, the dunning target determination unit 132 determines, as the dunning target T, the target debtor SA whose dunning scores, calculated by multiplying the dunning effectiveness identified by the effectiveness identification unit 131 by the debt balance, are equal to or greater than a threshold.

[0055] In the dunning classification table shown in Figure 5, the target debtor SA of SA009 and the target debtor SA of SA011 belong to the invalid dunning group, but have large debt balances, and are therefore classified as dunning target T. In this way, the dunning target determination unit 132 determines the dunning target T based on the debt balance of the target debtor SA, so that the target debtor SA with a large debt balance is more likely to be classified as dunning target T. As a result, the amount of receivables recovered through dunning operations can be maximized.

[0056] However, it is difficult to collect claims from target debtors SA that belong to the ineffective dunning group corresponding to the second validity data. Therefore, the dunning target determination unit 132 determines, for example, large debtors among the target debtors SA that correspond to the second validity data, whose debt balance is equal to or exceeds a threshold, as dunning targets T for whom dunning work will be performed by a dunning staff member P whose dunning ability value satisfies a predetermined condition. As an example, the dunning target determination unit 132 determines large debtors among the target debtors SA that belong to the ineffective dunning group, as dunning targets T for whom dunning work will be performed by a dunning staff member P whose dunning ability value is at the "expert level" in the dunning ability value table shown in Figure 4.

[0057] In this way, the dunning target determination unit 132 determines that an experienced dunning officer P will be responsible for dunning large debtors among the target debtors SA in the dunning invalid group. This increases the probability of collecting receivables from target debtors SA in the dunning invalid group compared to when an intermediate or beginner dunning officer P is responsible for dunning target debtors SA in the dunning invalid group. In addition, it is possible to reduce the loss of time and effort caused by intermediate or beginner dunning officers P failing to dunning, thereby maintaining the motivation of intermediate or beginner dunning officers P.

[0058] After the dunning target determination unit 132 determines the dunning target persons T, the dunning means determination unit 133 determines the dunning means for each dunning target person T based on the payment probability for each dunning means output by the payment prediction model for each dunning target person T. The dunning means determination unit 133 determines, as the dunning means to be executed for each dunning target person T, the dunning means for which the payment probability output by the payment prediction model, into which the attribute or effectiveness data output by the classification model of the dunning target person T has been input, is relatively high. For example, the dunning means determination unit 133 determines, as the dunning means to be executed, one or more dunning means for which the probability of payment is relatively high among the payment probabilities corresponding to manual calls, automatic calls, emails, SMS, letters, and postcards by dunning staff P output by the payment prediction model. The dunning means determination unit 133 may generate a dunning means table showing the payment probabilities for each dunning target person T by each dunning means.

[0059] 6 is a diagram showing an example of a dunning means table. In the dunning means table, the dunning target ID, the dunning means, the probability of payment for each dunning means, and a priority are associated with each other. A higher priority is assigned to a dunning means with a higher probability of payment. As will be described in detail later, the dunning means determination unit 133 can determine the number of times each dunning means should be executed by referring to the generated dunning means table.

[0060] Incidentally, when the probability of receiving a payment through a manual call by the dunning staff member P is about the same as the probability of receiving a payment through an automated call, it is preferable to employ an automated call in order to reduce the workload of the dunning staff member P. Therefore, when the difference between the probability of receiving a payment through a manual call by the dunning staff member P and the probability of receiving a payment through an automated call, as output by the payment prediction model, is less than a threshold value (for example, 10%), the dunning means determination unit 133 determines that an automated call will be the dunning means to be executed. This reduces the workload of the dunning staff member P.

[0061] The dunning means determination unit 133 may determine the dunning means for each dunning target person T based on the response probability for each dunning means output by the response probability prediction model for each dunning target person T. For example, the dunning means determination unit 133 determines the dunning means for each dunning target person T to be the dunning means to be executed, the dunning means having a relatively high response probability output by the response probability prediction model to which the attributes of the dunning target person T have been input.

[0062] As an example, the dunning means determination unit 133 may compare the response probability of dunning target T when a dunning call is made to the dunning target T's landline telephone, as output by the response probability prediction model, with the response probability of dunning target T when a dunning call is made to the dunning target T's mobile phone, and determine the dunning means with the higher probability as the dunning means to be executed. By doing so, the probability that dunning target T will respond to a dunning call increases, thereby improving the efficiency of dunning operations for debtors.

[0063] After determining the execution dunning means for each dunning target T in this way, the dunning means determination unit 133 determines the number of times the execution dunning means will be executed so as to maximize the total amount of receivables recovered through the dunning work. The expected value of the total amount of receivables recovered through the dunning work is the sum of the value obtained by multiplying the probability of deposit corresponding to the execution dunning means by the debt balance of the target debtor SA and the number of times the execution dunning means will be executed, for the number of dunning target T. The dunning means determination unit 133 assigns the total number of times dunning can be executed to the number of times the execution dunning means will be executed for each dunning target T so as to maximize this total value.

[0064] The dunning means determination unit 133 determines the number of times the execution dunning means is to be executed for each dunning target T shown in the dunning means table shown in Fig. 6, for example, so that the number of times the execution dunning means is to be executed for dunning target T with a larger debt balance is increased. This makes it possible to increase the expected value of the total amount to be collected. However, even for dunning target T with a small debt balance, although the priority of dunning is low, there are cases where the payment is executed with just one dunning, so it is desirable to execute at least one dunning.

[0065] Therefore, the dunning means determination unit 133 may determine the number of times the execution dunning means is to be executed for each dunning target T so that the number of times the execution dunning means is to be executed for all dunning target T is at least one. This prevents dunning from being completely omitted for dunning target T with a small debt balance, thereby reducing loss of debt collection. The dunning means determination unit 133 may generate a dunning execution count table that indicates the number of times each dunning means is executed for each dunning target T.

[0066] FIG. 7 is a diagram showing an example of a dunning execution count table. In the dunning execution count table, a dunning target ID, an execution dunning means, and the number of times each execution dunning means is executed are associated. As shown in FIG. 7, a higher dunning execution count is set for dunning target T T002, who has a relatively large debt balance. On the other hand, a dunning execution count of one is also set for dunning target T T001 and dunning target T T003, who have relatively small debt balances. The dunning means determination unit 133 stores the generated dunning execution count table in the memory unit 12. As a result, the schedule generation unit 135 can generate a dunning schedule by referring to the dunning execution count table, as will be described in detail later.

[0067] After the dunning target determination unit 132 determines the dunning target persons T, the dunning timing determination unit 134 determines the dunning timing for each dunning target person T based on the response probability for each dunning timing output by the response probability prediction model for each dunning target person T. The dunning timing determination unit 134 determines the timing for each dunning target person T at which the response probability output by the response probability prediction model to which the attributes of the dunning target person T have been input is relatively high.

[0068] As mentioned above, the response probability prediction model can also output the response probability to a reminder call to a landline and the response probability to a reminder call to a mobile phone. Therefore, the response probability prediction model may output the response probability for each combination (16 combinations in total) of reminder means (landline, mobile phone), date type (weekday, holiday), and time period (four time periods per day). In this case, the reminder timing determination unit 134 may generate a reminder timing table that indicates the response probability for each of the 16 combinations for each reminder target T.

[0069] 8 is a diagram showing an example of a reminder timing table, in which reminder target IDs, reminder means, date types, time periods, and response probabilities for each of 16 combinations are associated with each other.

[0070] The reminder timing determination unit 134 may determine one or more combinations of the 16 combinations with a relatively high response probability as a reminder combination, which is a combination of the reminder means to be executed and the reminder timing for the reminder target T. The number of combinations, which is one or more, is, for example, the number of reminder executions determined by the reminder target determination unit 132 for each reminder target T. In the example of FIG. 8, for the reminder target T of T001, who has been reminded once, the reminder timing determination unit 134 determines the combination with the highest response probability (the combination corresponding to the response probability shown in halftone dots) as the reminder combination. Furthermore, for the reminder target T of T002, who has been reminded three times, the reminder timing determination unit 134 determines the three combinations with the highest, second highest, and third highest response probabilities (the combinations corresponding to the three response probabilities shown in halftone dots) as the reminder combinations.

[0071] The reminder timing determination unit 134 stores the generated reminder timing table in the storage unit 12. As a result, the schedule generation unit 135 can generate a reminder schedule by referring to the reminder timing table, as will be described in detail later.

[0072] The dunning means determination unit 133 determines the dunning means to be executed and the number of dunning executions for each dunning target person T, and the dunning timing determination unit 134 determines the dunning timing for each dunning target person T, and then the schedule generation unit 135 generates a dunning schedule. The schedule generation unit 135 generates a dunning schedule by assigning dunning tasks for the dates and times corresponding to the dunning timing determined by the dunning timing determination unit 134 to the dunning staff P who are available to work on those dates and times. The schedule generation unit 135 generates a dunning schedule, for example, by assigning dunning tasks for the dunning timing determined by the dunning timing determination unit 134 to the dates and times when the dunning staff P, indicated in the availability status table stored in the memory unit 12, is available to perform the dunning tasks.

[0073] The schedule generation unit 135 may generate a dunning schedule by assigning dunning tasks to the dunning staff P in order of those dunning recipients T who tend to make deposits early. The schedule generation unit 135 generates a dunning schedule, for example, by assigning dunning tasks to the dunning staff P in order of those dunning recipients T with the shortest reference deposit period, which is the period from the dunning date when the dunning staff P made the dunning request to the deposit execution date when the dunning recipient T made the deposit. In this case, it is preferable that the reference dunning date be a dunning date within the last three months. Note that if there are multiple periods from the dunning date to the deposit execution date for one dunning recipient T, the schedule generation unit 135 may use the shortest period of the multiple periods as the reference deposit period for that dunning recipient T, or may use the average of the multiple periods as the reference deposit period for that dunning recipient T.

[0074] By doing this, the timing of deposit execution is earlier than when the dunning work is assigned to the dunning person P without prioritizing the dunning target T, making it possible to collect the receivables earlier.

[0075] The schedule generation unit 135 may generate a dunning schedule so that the number of dunning reminders per day for each dunning target T is limited to three, in order to ensure compliance with laws, regulations, and rules. This allows the dunning representative P to make dunning reminders in compliance with laws, regulations, and rules, thereby preventing any damage to the image of the financial institution that collects the debt. The process of generating a dunning schedule will be described in detail below.

[0076] 9 is a flowchart showing the flow of processing related to the generation of a dunning schedule. First, the schedule generation unit 135 determines whether the target debtor SA indicated in the debtor list data stored in the memory unit 12 corresponds to the dunning-free group by referring to the dunning classification table shown in FIG. 5 stored in the memory unit 12 (S51). If the target debtor SA corresponds to the dunning-free group (S51: YES), dunning for the target debtor SA is not necessary, and the processing ends.

[0077] On the other hand, if the target debtor SA does not fall into the dunning-unnecessary group (S51: NO), the schedule generation unit 135 determines whether the target debtor SA indicated by the debtor list data falls into the dunning-enabled group by referring to the dunning classification table shown in Fig. 5 stored in the storage unit 12 (S52). If the target debtor SA falls into the dunning-enabled group (S52: YES), the schedule generation unit 135 assigns an intermediate or junior level dunning officer P to the target debtor SA (S53).

[0078] On the other hand, if the target debtor SA does not fall into the valid dunning group (S52: NO), the target debtor SA falls into the invalid dunning group. In this case, the schedule generation unit 135 refers to the dunning execution count table shown in Figure 7 stored in the memory unit 12, and determines whether the dunning method for the target debtor SA, which belongs to the invalid dunning group, is other than a manual call (S54). If the dunning method for the target debtor SA, which belongs to the invalid dunning group, is other than a manual call (S54: YES), the schedule generation unit 135 assigns an intermediate-level or beginner-level dunning officer P to the target debtor SA (S53).

[0079] On the other hand, if the dunning method for the target debtor SA belonging to the invalid dunning group is manual call (S54: NO), a high level of verbal negotiation ability with the target debtor SA is required, so the schedule generation unit 135 assigns an experienced dunning staff member P to the target debtor SA (S55). By doing so, it is possible to effectively utilize the valuable human resource of an experienced dunning staff member P.

[0080] Through the processing up to this point, the type of dunning person P to be assigned to each target debtor SA (beginner-level or intermediate-level dunning person P, or experienced-level dunning person P) has been determined. Next, the schedule generation unit 135 assigns dunning timings with relatively high response probabilities, which are indicated in the dunning execution count table shown in Figure 7 stored in the memory unit 12, to the dunning means indicated in the dunning execution count table shown in Figure 7 stored in the memory unit 12 (S56).

[0081] As an example, for "Manual call: 1st execution" of target debtor SA of T001 in the dunning execution count table shown in Fig. 7, the schedule generation unit 135 assigns "Dunning method: mobile phone, date type: holiday, time period: 11:00-13:00", which has the highest response probability in the dunning timing table shown in Fig. 8. Furthermore, for "Manual call: 3rd execution" of target debtor SA of T002 in the dunning execution count table shown in Fig. 7, the schedule generation unit 135 assigns "Dunning method: landline, date type: weekday, time period: 11:00-13:00", "Dunning method: landline, date type: weekday, time period: 17:00-20:00", and "Dunning method: mobile phone, date type: holiday, time period: 13:00-17:00", which have relatively high response probabilities in the dunning timing table shown in Fig. 8.

[0082] Finally, the schedule generation unit 135 assigns, to the dunning timing assigned in S56, the dunning available date and time when the dunning person P, which is indicated in the availability status table stored in the storage unit 12, is available to perform the dunning work (S57). For example, the schedule generation unit 135 assigns the dunning available date and time of the beginner-level or intermediate-level dunning person P to the dunning timing for the dunning target person T to whom a beginner-level or intermediate-level dunning person P is assigned in S53. Furthermore, for example, the schedule generation unit 135 assigns the dunning available date and time of the skilled-level dunning person P to the dunning timing for the dunning target person T to whom a skilled-level dunning person P is assigned in S55. In this way, the schedule generation unit 135 can generate a dunning schedule table.

[0083] 10 is a diagram showing an example of a dunning schedule table. In the dunning schedule table, a dunning person ID, a dunning target person ID, a dunning method, and a dunning date and time are associated with each other. The schedule generation unit 135 transmits the generated dunning schedule table to the information terminal 2 used by the dunning person P. This allows the dunning person P to carry out dunning by referring to the generated dunning schedule.

[0084] [Effects of information processing device 1] As explained above, the information processing device 1 identifies the dunning effectiveness for each of multiple target debtors SA based on the effectiveness data output by the classification model to which the attributes of the target debtors SA to be classified are input, and determines some target debtors SA with relatively high dunning effectiveness as dunning targets T. As a result, the number of debtors to be dunned is narrowed down, thereby improving the efficiency of dunning operations.

[0085] Furthermore, the information processing device 1 generates a dunning schedule indicating the schedule of dunning to be carried out by the dunning person P. By referring to the generated dunning schedule, the dunning person P can easily understand which dunning means to use and when to dunning to which dunning target person T. As a result, the efficiency of the dunning work can be further improved.

[0086] <Modification> In the above embodiment, the effectiveness identification unit 131 identified the dunning effectiveness for each of the multiple target debtors SA based on the effectiveness data output by the classification model, which is a machine learning model. In a modified example, the effectiveness identification unit 131 identifies the dunning effectiveness for each of the multiple target debtors SA based on an effectiveness table rather than a machine learning model. In the modified example, the memory unit 12 stores an effectiveness table in which debtor attributes are associated with effectiveness data indicating the effectiveness of dunning to encourage the debtor to make payment.

[0087] Fig. 11 is a diagram showing an example of an effectiveness table. In the effectiveness table, the attributes of the debtor are associated with effectiveness data indicating the effectiveness of the reminder. For the effectiveness data in Fig. 11, "unnecessary" indicates the first effectiveness data, "invalid" indicates the second effectiveness data, and "effective" indicates the third effectiveness data.

[0088] The effectiveness specification unit 131 specifies the dunning effectiveness for each of multiple target debtors SA based on effectiveness data associated in the effectiveness table with the attributes of debtors that have a similarity equal to or greater than a predetermined threshold with the attributes of the target debtor SA to be classified. The effectiveness specification unit 131 specifies the dunning effectiveness for each of multiple target debtors SA so that, for example, the dunning effectiveness of a target debtor SA whose effectiveness data is "effective" is higher than the dunning effectiveness of a target debtor SA whose effectiveness data is "unnecessary" and the dunning effectiveness of a target debtor SA whose effectiveness data is "invalid."

[0089] From this point on, the processes executed by the dunning target determination unit 132, dunning means determination unit 133, dunning timing determination unit 134, and schedule generation unit 135 are the same as those explained in the above embodiment, and therefore explanations thereof will be omitted.

[0090] The present invention has been described above using embodiments, but the technical scope of the present invention is not limited to the scope described in the above embodiments, and various modifications and changes are possible within the scope of the gist of the present invention. For example, all or part of the device can be configured by functionally or physically distributing or integrating any unit. Furthermore, new embodiments resulting from any combination of multiple embodiments are also included in the embodiments of the present invention. The effects of the new embodiments resulting from the combination also have the effects of the original embodiments. [Explanation of symbols]

[0091] 1. Information processing equipment 11 Communications Department 12 Storage section 13 Control Unit 131 Effectiveness Identification Unit 132 Collection Target Determination Division 133 Dunning Means Determination Department 134 Reminder Timing Determination Department 135 Schedule Generation Unit 2. Information terminal S Information Processing System

Claims

1. a memory unit that stores a classification model that is machine-learned using training data consisting of multiple sets of debtor attributes and effectiveness data indicating the effectiveness of reminders to encourage the debtors to make payments, and that, when the debtor attributes are input, outputs, in association with each of the multiple debtors, one of the following effectiveness data: first effectiveness data indicating that the probability of making a payment without reminders is equal to or greater than a first threshold, second effectiveness data indicating that the probability of making a payment after reminders is less than a second threshold that is smaller than the first threshold, or third effectiveness data indicating that the probability of making a payment without reminders is less than the first threshold and that the probability of making a payment after reminders is equal to or greater than the second threshold; an effectiveness determination unit that determines the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified are input; a dunning target determination unit that determines, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; An information processing device having the above.

2. a memory unit that stores a machine learning model that is machine-learned using multiple sets of debtor attributes and effectiveness data that indicate the effectiveness of dunning to encourage the debtor to make a payment as training data, the machine learning model outputting the effectiveness data when the debtor attributes are input, and a payment prediction model outputting the probability of payment by dunning means when the debtor attributes or the effectiveness data are input; an effectiveness determination unit that determines the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified are input; a dunning target determination unit that determines, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; a dunning means determination unit that determines the dunning means for which the probability of receiving payment output by the payment prediction model to which the attributes of the dunning target person or the output validity data has been input is relatively high as the dunning means to be executed for each of the dunning target persons; and The dunning means determination unit allocates the total number of times that the dunning can be carried out to the number of times that the execution dunning means is executed for each of the dunning targets so that the total value obtained by multiplying the probability of payment corresponding to the execution dunning means for each of the dunning targets by the debt balance of the target debtor and the number of times that the execution dunning means is executed is maximized for the number of dunning targets. Information processing device.

3. a memory unit that stores a machine learning model that is machine-learned using multiple sets of debtor attributes and effectiveness data that indicate the effectiveness of dunning to encourage the debtor to make a payment as training data, the machine learning model outputting the effectiveness data when the debtor attributes are input, and a payment prediction model outputting the probability of payment by dunning means when the debtor attributes or the effectiveness data are input; an effectiveness determination unit that determines the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified are input; a dunning target determination unit that determines, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; a dunning means determination unit that determines the dunning means for which the probability of receiving payment output by the payment prediction model to which the attributes of the dunning target person or the output validity data has been input is relatively high as the dunning means to be executed for each of the dunning target persons; and The dunning means determination unit determines the automatic call as the execution dunning means when the difference between the probability of receiving a payment by a manual call by a dunning person and the probability of receiving a payment by an automatic call, as output by the payment prediction model, is less than a threshold value. Information processing device.

4. the dunning target determination unit determines the target debtor corresponding to the third validity data as the dunning target; The information processing device according to claim 1 .

5. the storage unit stores a dunning ability value indicating the ability of a dunning person to perform dunning work; the dunning target determination unit determines, among the target debtors corresponding to the second validity data, those whose debt balance is equal to or greater than a threshold, as dunning targets for which the dunning staff, whose dunning ability value satisfies a predetermined condition, will perform the dunning work; The information processing device according to claim 1 .

6. the dunning target determination unit determines, among the plurality of target debtors, some of the target debtors having relatively high dunning scores calculated based on the dunning effectiveness and the debt balances of the debtors as the dunning targets; The information processing device according to claim 1 .

7. the storage unit further stores a response probability prediction model that, when the attributes of the debtor are input, outputs a response probability of the debtor to the reminder for each reminder timing; the information processing device further includes a reminder timing determination unit that determines the timing at which the response probability output from the response probability prediction model to which attributes of the reminder recipients are input is relatively high as the reminder timing for each of the reminder recipients; The information processing device according to claim 1 .

8. The storage unit stores the availability status of the dunning staff by date and time, The information processing device further includes a schedule generation unit that generates a dunning schedule by assigning a dunning task for a date and time corresponding to the dunning timing to the dunning staff who are available to work on the date and time. The information processing device according to claim 7 .

9. the schedule generation unit generates the dunning schedule by assigning to the dunning staff the dunning tasks for the dunning target person in order of the period from the dunning date made by the dunning staff to the deposit execution date made by the dunning target person, The information processing device according to claim 8 .

10. the storage unit further stores a response probability prediction model that outputs a response probability of the debtor to the dunning for each dunning means when the attributes of the debtor are input; The information processing device further includes a dunning means determination unit that determines the dunning means for which the response probability output by the response probability prediction model to which attributes of the dunning target persons are input is relatively high as the execution dunning means for each of the dunning target persons. The information processing device according to claim 1 .

11. a storage unit that stores an effectiveness table in which attributes of a debtor are associated with effectiveness data that indicates the effectiveness of a reminder to the debtor to make a payment, the effectiveness data being one of the following: first effectiveness data indicating that the probability of the debtor making a payment without a reminder is equal to or greater than a first threshold, second effectiveness data indicating that the probability of the debtor making a payment after a reminder is less than a second threshold that is smaller than the first threshold, or third effectiveness data indicating that the probability of the debtor making a payment without a reminder is less than the first threshold and that the probability of the debtor making a payment after a reminder is equal to or greater than the second threshold; an effectiveness specification unit that specifies the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data associated in the effectiveness table with the attributes of the debtors that have a similarity equal to or greater than a predetermined threshold value with the attributes of the target debtors to be classified; a dunning target determination unit that determines, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; An information processing device having the above.

12. A machine learning model that has been machine-learned using multiple sets of debtor attributes and effectiveness data indicating the effectiveness of reminders to encourage the debtors to make payments as training data, wherein when the debtor attributes are input, the machine learning model is executed by a computer having a storage unit that stores a classification model that, when inputted with each of the multiple debtors, outputs one of the following effectiveness data: first effectiveness data indicating that the probability of making a payment without reminders is equal to or greater than a first threshold, second effectiveness data indicating that the probability of making a payment after reminders is less than a second threshold that is smaller than the first threshold, or third effectiveness data indicating that the probability of making a payment without reminders is less than the first threshold and that the probability of making a payment after reminders is equal to or greater than the second threshold. an effectiveness determination step of determining the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified have been input; a dunning target determination step of determining, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; An information processing method comprising:

13. A machine learning model that is machine-learned using multiple sets of debtor attributes and effectiveness data indicating the effectiveness of dunning to encourage the debtor to make a payment as training data, the machine learning model being executed by a computer having a storage unit that stores: a classification model that outputs the effectiveness data when the debtor attributes are input; and a payment prediction model that outputs the probability of payment by dunning method when the debtor attributes or the effectiveness data are input. an effectiveness determination step of determining the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified have been input; a dunning target determination step of determining, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; a dunning means determination step of determining the dunning means for which the probability of receiving payment output by the payment prediction model into which the attributes of the dunning target person or the output validity data has been input is relatively high as the dunning means to be executed for each of the dunning target people; and In the dunning means determination step, the total number of times that the dunning can be carried out is assigned to the number of times that the execution dunning means for each of the dunning targets is executed so that the total value obtained by multiplying the payment probability corresponding to the execution dunning means for each of the dunning targets by the debt balance of the target debtor and the number of times that the execution dunning means is executed is maximized. Information processing methods.

14. A machine learning model that is machine-learned using multiple sets of debtor attributes and effectiveness data indicating the effectiveness of dunning to encourage the debtor to make a payment as training data, the machine learning model being executed by a computer having a storage unit that stores: a classification model that outputs the effectiveness data when the debtor attributes are input; and a payment prediction model that outputs the probability of payment by dunning method when the debtor attributes or the effectiveness data are input. an effectiveness determination step of determining the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified have been input; a dunning target determination step of determining, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; a dunning means determination step of determining the dunning means for which the probability of receiving payment output by the payment prediction model into which the attributes of the dunning target person or the output validity data has been input is relatively high as the dunning means to be executed for each of the dunning target people; and In the dunning means determination step, if the difference between the probability of receiving a payment by a manual call by a dunning staff member and the probability of receiving a payment by an automatic call, as output by the payment prediction model, is less than a threshold value, the automatic call is determined to be the dunning means to be executed. Information processing methods.

15. a processor having a memory unit for storing a classification model that is machine-learned using as training data a plurality of sets of debtor attributes and effectiveness data indicating the effectiveness of reminders to encourage the debtors to make payments, the classification model being trained by the processor when the debtor attributes are input, and that outputs one of the following effectiveness data associated with each of the plurality of debtors: first effectiveness data indicating that the probability of making a payment without reminders is equal to or greater than a first threshold, second effectiveness data indicating that the probability of making a payment after reminders is less than a second threshold that is smaller than the first threshold, or third effectiveness data indicating that the probability of making a payment without reminders is less than the first threshold and that the probability of making a payment after reminders is equal to or greater than the second threshold; an effectiveness determination unit that determines the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified are input; a dunning target determination unit that determines, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; A program to make it function as such.

16. A machine learning model is machine-learned using training data consisting of multiple sets of debtor attributes and effectiveness data indicating the effectiveness of dunning to encourage the debtor to make a payment, the machine learning model having a memory unit that stores: a classification model that outputs the effectiveness data when the debtor attributes are input; and a payment prediction model that outputs the probability of payment by dunning method when the debtor attributes or the effectiveness data are input; an effectiveness determination unit that determines the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified are input; a dunning target determination unit that determines, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; a dunning means determination unit that determines the dunning means for which the probability of receiving payment output by the payment prediction model to which the attributes of the dunning target person or the output validity data has been input is relatively high as the dunning means to be executed for each of the dunning target persons; A program for making the device function as a The dunning means determination unit allocates the total number of times that the dunning can be carried out to the number of times that the execution dunning means is executed for each of the dunning targets so that the total value obtained by multiplying the probability of payment corresponding to the execution dunning means for each of the dunning targets by the debt balance of the target debtor and the number of times that the execution dunning means is executed is maximized for the number of dunning targets. program.

17. A machine learning model is machine-learned using training data consisting of multiple sets of debtor attributes and effectiveness data indicating the effectiveness of dunning to encourage the debtor to make a payment, the machine learning model having a memory unit that stores: a classification model that outputs the effectiveness data when the debtor attributes are input; and a payment prediction model that outputs the probability of payment by dunning method when the debtor attributes or the effectiveness data are input; an effectiveness determination unit that determines the effectiveness of dunning for each of the plurality of target debtors based on the effectiveness data output by the classification model to which attributes of the target debtors to be classified are input; a dunning target determination unit that determines, among the plurality of target debtors, some of the target debtors having a relatively high dunning effectiveness as dunning targets; a dunning means determination unit that determines the dunning means for which the probability of receiving payment output by the payment prediction model to which the attributes of the dunning target person or the output validity data has been input is relatively high as the dunning means to be executed for each of the dunning target persons; A program for making the device function as a The dunning means determination unit determines the automatic call as the execution dunning means when the difference between the probability of receiving a payment by a manual call by a dunning person and the probability of receiving a payment by an automatic call, as output by the payment prediction model, is less than a threshold value. program.

Images