JP7194309B1 - Information processing system, program and information processing method - Google Patents

Abstract

The object of the present invention is to improve the efficiency of dunning for the obligor of a bond by using machine learning. An information processing system uses a learning model to make a prediction as to whether a receivable will be paid by the obligee if or without dunning. It has a control unit that performs The learning model uses explanatory variable data for each of the multiple claims and whether payment was made as a result of performing a dunning action or not as a result of not performing a dunning action for each obligor of the multiple claims. and objective variable data representing whether or not payment is made, which indicates whether or not payment has been made. The control unit acquires target bond data for a prediction target bond. By inputting the target receivables data into the learning model, the obligors of the receivables to be predicted will pay the receivables to be predicted if they perform a dunning action or if they do not perform a dunning action. or make a prediction. [Selection drawing] Fig. 6

Description

The present invention relates to an information processing system, a program, and an information processing method.

In the prior art, the obligor characteristic factor element is determined based on the contract date, promised date, and payment date, and is added to the accumulated obligor characteristic factor element, which is a set of the previous obligor characteristic factor elements, and the added cumulative debt The customer type that best matches the characteristics of the debtor is selected based on the borrower characteristics factor element, and priority is given to output the demanding telephone service, so that the operator can carry out the telephone service that matches the customer type. There is known a technology capable of performing efficient dunning work (Patent Literature 1).

JP 2007-094998 A

However, Patent Literature 1 does not disclose that the prompting is made more efficient using machine learning.

In view of the above circumstances, it is an object of the present invention to provide a technique for making reminders more efficient using machine learning.

According to one aspect of the invention, an information processing system is provided. This information processing system is an information processing system that uses a learning model to determine whether the debtor will make a payment to the debtor if the demanding action is performed or if the demanding action is not performed. It has a control unit that makes predictions about The learning model uses explanatory variable data for each of the multiple claims and whether payment was made as a result of performing a dunning action or not as a result of not performing a dunning action for each obligor of the multiple claims. The learning model obtained by learning the objective variable data representing the result of whether or not payment was made, and the result of whether or not payment was made. The control unit acquires target bond data for a prediction target bond. By inputting the target receivables data into the learning model, the obligors of the receivables to be predicted will make payments to the receivables to be predicted if they perform a dunning action or if they do not perform a dunning action. make predictions about whether

It is a figure explaining an example of a business model in this indication. It is a figure which shows an example of the type of claim in this disclosure. 1 is a diagram showing an example of a system configuration of an information processing system 1 according to Embodiment 1; FIG. 2 is a diagram illustrating an example of a hardware configuration of a server device 2 according to Embodiment 1; FIG. 3 is a diagram illustrating an example of a hardware configuration of a client device 3 according to Embodiment 1; FIG. 3 is a diagram showing an example of an activity showing the flow of information processing executed by the information processing system 1 according to the first embodiment; FIG. FIG. 5 is a diagram showing an example of a process of performing machine learning on the learning model 5 according to the first embodiment; 4 is a diagram showing an example of explanatory variable data 40 according to the first embodiment; FIG. FIG. 7 is a diagram showing an example of a process of making predictions using a learned learning model 7 according to the first embodiment; FIG. 10 is a diagram showing an example of a telephone demanding payment data table T1 according to the first embodiment; FIG. FIG. 10 is a diagram showing an example of a telephone reminder unpaid data table T2 according to the second embodiment; FIG. 13 is a diagram showing an example of a natural payment data table T3 according to Embodiment 3; FIG. 14 is a diagram showing an example of a bond prediction data table T4 according to Embodiment 4;

[Embodiment]
Embodiments of the present invention will be described below with reference to the drawings. Various features shown in the embodiments shown below can be combined with each other.

In the present embodiment, an example of collection of claims based on the business model shown in FIG. 1 will be described. FIG. 1 is a diagram illustrating an example of a business model in the present disclosure. In FIG. 1, seller X is the seller of product P. Debtor Y is the buyer of product P. Debtor Y is the obligee of the receivable relating to the payment of product P. The agent Z purchases the receivables from the seller X and collects the receivables related to the payment of the product P on behalf of the seller X. Agent Z is the obligee of the claim relating to the payment of product P.

In the business model shown in FIG. 1, seller X first sells product P to debtor Y on credit. The seller X, who has sold the product P on credit, notifies the agent Z of a request for collection of the credit of the product P by the notification means N. The agent Z, who has been contacted by the seller X by means of notification N, examines the claim in consideration of the risk of bad debt. The agent Z pays the seller X the price M1 for the product P for the claim that has passed the examination. Here, the commission of the agent Z may be deducted from the sales amount of the product P to obtain the price M1. In addition, the agent Z requests the debtor Y to pay the claim for the product P using the claiming means I. The obligor Y, who has been billed by the billing means I, pays the agent Z the price M2. The above is a series of flows of this business model. This business model is also called factoring. It should be noted that the same applies to the case where a service-related transaction is performed instead of the product P.

Some obligors Y have delayed payments by not making payments on their receivables on time. Agent Z demands payment from obligor Y of such a claim by e-mail, telephone, visit, or the like.

In this embodiment, bonds are classified into the types shown in FIG. FIG. 2 is a diagram showing an example of types of claims in the present disclosure. That is, the claims are at least classified into claims for which payment is made within the payment deadline and claims for which payment is not made within the payment deadline. In addition, receivables whose payment is not made within the due date are divided into receivables that are paid naturally after the due date (class A) and receivables that are paid when demanded (class B), and receivables that are not paid even if demanded. receivables (hereinafter Class C). Regarding the determination of whether or not payment was made despite the demand for classification into Class C, if the payment has already been demanded and a certain period of time has elapsed, it is judged that the payment was not made despite the demand. may be From the viewpoint of cost, time, human resources, etc., agent Z wants to demand payment efficiently without calling or visiting as much as possible. For example, it is not considered necessary to implement demanding actions for claims that fall under Class A, but there is no significance in implementing demanding actions for claims that fall under Class B. This is because it is considered to be high.

In this embodiment, in factoring, receivables for which payment is not made within the deadline are receivables that will be paid naturally even though the deadline has passed, receivables that will be paid if demanded, or receivables that will not be paid even if demanded. Explain how to reasonably predict whether or not there is a claim.

By the way, the program for realizing the software appearing in this embodiment may be provided as a non-transitory computer-readable medium (Non-Transitory Computer-Readable Medium), or may be downloaded from an external server. It may be provided as possible, or may be provided so that the program is activated by an external computer and the function is realized by the client device (so-called cloud computing).

Further, in the present embodiment, the term “unit” may include, for example, a combination of hardware resources implemented by circuits in a broad sense and software information processing that can be specifically realized by these hardware resources. . In addition, various information is handled in this embodiment, and these information are, for example, physical values of signal values representing voltage and current, and signal values as binary bit aggregates composed of 0 or 1. , or by quantum superposition (so-called quantum bits), and communications and operations can be performed on circuits in a broad sense.

A circuit in a broad sense is a circuit implemented by appropriately combining at least circuits, circuits, processors, memories, and the like. Application Specific Integrated Circuits (ASICs), programmable logic devices (e.g., Simple Programmable Logic Devices (SPLDs), Complex Programmable Logic Devices (CPLDs), and field It includes a programmable gate array (Field Programmable Gate Array: FPGA).

[Embodiment 1]
1. System Configuration of Information Processing System 1 First, the system configuration of the information processing system 1 of the first embodiment will be described with reference to FIG.

FIG. 3 is a diagram showing an example of the system configuration of the information processing system 1 according to the first embodiment. As shown in FIG. 3, the information processing system 1 includes a server device 2, a client device 3, and a network NW. The server device 2 is configured to be able to communicate with the client device 3 via the network NW. As a result, the server device 2 and the client device 3 can mutually transmit or receive various information. Note that the server device 2 and the client device 3 are examples of information processing devices, and are not limited to the present embodiment. The client device 3 may be a PC (Personal Computer), a tablet computer, a smart phone, or the like. Note that there may be a plurality of server devices 2 and client devices 3 .

2. Hardware Configuration Next, the hardware configuration of the server apparatus 2 according to the first embodiment will be described with reference to FIG. 4, and the hardware configuration of the client apparatus 3 according to the first embodiment will be described with reference to FIG.

2.1. Hardware Configuration of Server Apparatus 2 FIG. 4 is a diagram illustrating an example of a hardware configuration of the server apparatus 2 according to the first embodiment. As shown in FIG. 4, the server device 2 includes a control unit 21, a storage unit 22, and a communication unit 23. These components are electrically connected inside the server device 2 via a communication bus. It is The server device 2 executes processing according to the embodiment.

The control unit 21 processes and controls overall operations related to the server device 2 . The control unit 21 is, for example, a central processing unit (CPU). By the control unit 21 reading out a predetermined program stored in the storage unit 22 and executing processing based on the program, various functions related to the server device 2, such as processing shown in FIGS. is realized. For example, the control unit 21 uses the learning model to predict whether the debtor will pay the debt if the debtor of the debt is dunned or not. Note that the control unit 21 is not limited to a single unit, and a plurality of control units 21 may be provided for each function. A combination thereof may also be used.

The storage unit 22 stores various information defined by the above description. For example, as a storage device such as a solid state drive (SSD) that stores various programs related to the server device 2 executed by the control unit 21, a temporary necessary storage device related to the calculation of the program is used. It may be implemented as a memory such as a Random Access Memory (RAM) for storing information (arguments, arrays, etc.). The storage unit 22 stores various programs related to the server device 2 executed by the control unit 21, variables, data used when the control unit 21 executes processing based on the programs, and the like. The storage unit 22 is an example of a storage medium.

The communication unit 23 preferably uses wired communication means such as USB, IEEE1394, Thunderbolt (registered trademark), wired LAN network communication, etc., but wireless LAN network communication, mobile communication such as LTE/3G/4G/5G, BLUETOOTH (registered trademark), etc. Trademark) communications, etc. may be included as appropriate. That is, it is more preferable to implement as a set of these communication means. That is, the server device 2 may communicate various information from the outside via the communication unit 23 .

2.2. Hardware Configuration of Client Apparatus 3 FIG. 5 is a diagram showing an example of the hardware configuration of the client apparatus 3 according to the first embodiment. As shown in FIG. 5, the client device 3 has a control unit 31, a storage unit 32, a communication unit 33, an input unit 34, and an output unit 35. These components are the components of the client device 3. They are electrically connected internally via a communication bus. The client device 3 executes processing according to the embodiment. The client device 3 is used by a user associated with agent Z. For example, the client device 3 is used by agent Z if agent Z is a natural person. Also, if the agent Z is a corporation, the client device 3 is used by an employee of the agent Z, an employee of an affiliated company of the agent Z, a person entrusted with work by the agent Z, or the like. For the control unit 31 , storage unit 32 and communication unit 33 of the client device 3 , refer to the control unit 21 , storage unit 22 and communication unit 23 of the server device 2 .

The input unit 34 may be included in the housing of the client device 3 or may be externally attached. For example, the input unit 34 may be integrated with the output unit 35 and implemented as a touch panel. A touch panel allows the user to input a tap operation, a swipe operation, or the like. Of course, a switch button, a mouse, a QWERTY keyboard, or the like may be employed instead of the touch panel. That is, the input unit 34 receives input based on the operation performed by the user. The input is transferred as a command signal to the control unit 31 via the communication bus, and the control unit 31 can perform predetermined control or calculation as necessary.

The output unit 35 can function as a display unit of the client device 3 . For example, the output unit 35 may be included in the housing of the client device 3 or may be externally attached. The output unit 35 displays a screen of a graphical user interface (GUI) that can be operated by the user. For example, it is preferable to use a display device such as a CRT display, a liquid crystal display, an organic EL display, and a plasma display according to the type of the client device 3 .

3. Information Processing of Information Processing System 1 Next, an example of preferable information processing executed by the information processing system 1 of the present embodiment will be described.

In this specification, various data transmitted and received via the network NW, the communication unit 23 and the communication unit 33 are stored in the storage unit 22 and the storage unit 32, respectively.

Also, in this specification, information processing related to the network NW, the communication unit 23, the communication unit 33, and the input unit 34 may be omitted. For example, the control unit 31 receiving a predetermined input through the user's operation on the input unit 34 may simply be described as the control unit 31 receiving a predetermined input from the user. In addition, the control unit 21 simply transmits information to the client device 3, or In some cases, it is simply described that the control unit 21 receives information from the client device 3 . The same applies when the control unit 31 transmits/receives information to/from the server device 2 via the communication unit 33, the network NW, and the communication unit 23. FIG.

In this specification, the information processing for displaying information on the output unit 35 is simply assumed to be information displayed by the control unit 31 on the output unit 35, and indirectly by displaying information by the control unit 21 on the output unit 35. may be abbreviated as what is displayed in the In that case, for example, information processing as described below is executed. For example, the control unit 31 receives an instruction from the user to change the state of the screen. Control unit 31 generates display information based on the instruction. The control unit 31 controls display on the output unit 35 based on the generated display information. As another example, the control unit 31 receives an instruction from the user to display information on the screen. Control unit 31 transmits the instruction to server device 2 . The control unit 21 receives the instruction from the client device 3 . Control unit 21 generates display information based on the instruction. The control unit 21 transmits the generated display information to the client device 3 . The control unit 31 receives the generated display information from the server device 2 . The control unit 31 controls display on the output unit 35 based on the generated display information. Here, the display information is a concept that includes not only information itself generated in a user-visible manner such as screens, images, icons, texts, etc., but also rendering information for display.

3.1 Outline of Information Processing In this section, an information processing method by the information processing system 1 described above will be described. FIG. 6 is a diagram showing an example of an activity showing the flow of information processing executed by the information processing system 1 according to the first embodiment. In FIG. 6, processes A1 to A6 are processes related to acquisition of the learned learning model 7, and processes A7 to A13 are processes related to prediction using the learned learning model 7. FIG.

First, the control unit 31 acquires teacher data 4a, 4b, and 4c including explanatory variable data 40 and objective variable data (A1). For example, the control unit 31 identifies, as the explanatory variable data 40, the data of the credit of the debtor who does not pay within the time limit shown in FIG. The control unit 31 acquires the teacher data 4a, 4b, and 4c by accepting the classification of objective variable data for each explanatory variable data 40 from the user.

Subsequently, the control unit 31 receives an instruction to start machine learning from the user. When receiving an instruction to start machine learning from the user, the control unit 31 transmits teacher data 4a, 4b, 4c and an instruction for machine learning to the server device 2 (A2).

Subsequently, the control unit 21 receives the teacher data 4a, 4b, 4c and machine learning instructions from the client device 3 (A3).

Subsequently, the control unit 21 uses the teaching data 4a, 4b, and 4c to cause the learning model 5 to perform machine learning using the explanatory variable data 40 as the explanatory variable and the objective variable data as the objective variable (A4). After that, the control unit 21 acquires the learning model 7 that has been learned. A4 will be described in detail with reference to FIG.

Subsequently, the control unit 21 transmits a notification to the effect that the trained learning model 7 has been acquired to the client device 3 (A5).

Subsequently, the control unit 31 receives a notification that the learned learning model 7 has been acquired from the server device 2 (A6). After that, the control unit 31 causes the output unit 35 to display a notification that the learned learning model 7 has been acquired. For example, the control unit 31 causes the output unit 35 to display information such as "the machine learning has been completed and the learning model has been obtained."

The processing of A1 to A6 above is the processing related to acquisition of the learning model 7 that has been trained. In FIG. 6, the processes A1 to A13 are described as if they were a series of processes, but the information processing system 1 includes the processes A1 to A6 related to acquisition of the trained learning model 7, and the processes A7 The processing related to prediction using the trained learning model 7 of A13 may be executed at different timings as a non-series processing. For example, if a trained model with high accuracy has already been acquired, and if learning results in a trained model with lower accuracy than before, the information processing system 1 performs the processing of A1 to A6. are omitted, and only the processing of A7 to A13 is executed.

Subsequently, the control unit 31 acquires the target bond data 6, which is data about the prediction target bond (A7). The control unit 31 receives a prediction instruction using the learned learning model 7 from the user.

Subsequently, when receiving a prediction instruction from the user, the control unit 31 transmits the target bond data 6 and the prediction instruction to the server device 2 (A8).

Subsequently, the control unit 31 receives the target bond data 6 and the prediction instruction from the client device 3 (A9).

Subsequently, the control unit 21 inputs the target bond data 6 to the learned learning model 7, and predicts whether or not the bond will be paid according to the presence or absence of the demanding action for the debtor (A10). For example, the control unit 21 predicts the probability of payment of the claim based on the presence or absence of a dunning action against the debtor. After that, the control unit 21 outputs output data 8 regarding the result of prediction as to whether or not the debtor will be paid, depending on the presence or absence of demanding action to the debtor. According to one aspect, the control unit 21 acquires the target bond data 6 for the prediction target bond. By inputting the target bond data 6 to the learned learning model 7, the control unit 21 instructs the obligor of the target bond to predict whether or not the obligor performs the dunning action. A prediction is made as to whether payment will be made on the predicted receivable. A10 will be described in detail with reference to FIG. Note that the control unit 21 may perform the prediction at regular time intervals, such as once a month, and the plurality of claims to be predicted at the timing of the prediction are all past due payments. There may be.

Subsequently, the control unit 21 transmits the output data 8 regarding the result of prediction to the client device 3 (A11).

Subsequently, the control unit 31 receives the output data 8 regarding the result of prediction from the server device 2 (A12).

Finally, the control unit 31 causes the output unit 35 to display the output data 8 regarding the result of prediction (A13). The output data 8 may include a telephone dunning payment data table T1. A13 will be described in detail with reference to FIG.

As described above, according to the present disclosure, it is possible to provide a technique for making reminders more efficient using machine learning. Further, in Patent Document 1, as shown in FIG. 8 of Patent Document 1, customer characteristics are determined based on whether the customer actually paid within the deadline, the number of times the customer made a phone call to remind the customer, and so on. and classified. Therefore, in Patent Literature 1, new customers are classified into the characteristic of "not yet known (initial stage)".

Further, in Patent Document 1, it is considered that a reliable classification cannot be achieved unless a history of payment actions and dunning actions by the same customer is accumulated. In addition, in the factoring targeted by this embodiment, demand for invoices with a seller X and a debtor Y is targeted, and there is also a difference in the business model from Patent Document 1.

According to the present disclosure, by using a learning model obtained by machine learning, for a new claim (new claim of debtor Y, or a new pair of claim of seller X and debtor Y) can also make predictions about the need for dunning. Furthermore, according to the present disclosure, it is possible to provide a technique for judging the necessity of prompting in a highly accurate and reliable manner using a learning model obtained by machine learning.

In addition, since complicated information processing is not required in performing processing using machine learning, cache memory usage of the server device 2 and the client device 3 can be reduced. In addition, as a result of being able to reduce the use of cache memory, a large-scale device or computer is not required, so information processing can be performed at low cost.

3.2 Details of Information Processing Next, details of the information processing outlined above will be described with reference to FIGS. 7 to 10. FIG. First, with reference to FIG. 7, the details of the machine learning process for the learning model 5 of activity A4 in FIG. 6 will be described. FIG. 7 is a diagram illustrating an example of a process of performing machine learning on the learning model 5 according to the first embodiment.

FIG. 7 includes a plurality of teacher data 4a, 4b, 4c and a learning model 5. FIG. The teacher data 4a, 4b, and 4c are data selected from the data of claims of debtors who do not pay within the time limit shown in FIG. The plurality of teacher data 4a, 4b, and 4c are data aggregated at arbitrary timing and for arbitrary periods and used as teacher data. The arbitrary timing is, for example, the 1st day of every month, every Monday, 50 days after the previous totalization, or the like. The arbitrary period is, for example, a period of 30 days from the latest aggregation, a period of 1 year from the latest aggregation, or the like.

The teacher data 4 a includes explanatory variable data 40 and natural payment data 41 . The explanatory variable data 40 will be described in detail using FIG. The spontaneous payment data 41 is data relating to whether or not the debt was paid by the debtor if the debtor did not make a demanding action by telephone after the payment deadline. The natural payment data 41 is an example of objective variable data. In addition to telephone calls, demanding acts such as e-mails and visits may also be performed, in which case the demanding natural payment data is used instead of the natural payment data 41 . The dunning natural payment data is data relating to whether or not the debt was paid by the debtor when the dunning act was not performed to the debtor after the payment deadline passed. The demanding natural payment data is an example of objective variable data.

The teacher data 4b includes explanatory variable data 40 and telephone reminder payment data 42. FIG. The telephone demanding payment data 42 is data relating to whether or not the debtor has paid the claim when the demanding action is executed by telephone to the debtor after the payment deadline has passed. The telephone dunning payment data 42 is an example of objective variable data. In addition to telephone calls, demanding actions such as e-mails and visits may be performed, in which case the demanding payment data is used instead of the telephone demanding payment data 42. The dunning payment data is data relating to whether or not the debtor has paid the claim when the dunning action is executed against the debtor after the payment deadline has passed. Demanding payment data is an example of objective variable data.

Some teacher data 4c includes explanatory variable data 40 and unpaid telephone reminder data 43. FIG. The unpaid telephone demand data 43 is data relating to whether or not the debtor has not paid the debt when the debtor is demanded by telephone after the payment deadline has passed. The telephone reminder unpaid data 43 is an example of objective variable data. In addition to telephone calls, reminder actions such as e-mails and visits may be performed, in which case unpaid reminder data is used instead of telephone reminder unpaid data 43 . The dunning unpaid data is data relating to whether or not the debtor did not pay the claim when the dunning action was executed to the debtor after the payment deadline passed. The demand unpaid data is an example of objective variable data.

Also, for example, teacher data 4a, 4b, and 4c are created as follows. The control unit 31 acquires the explanatory variable data 40 from the bill-related data. The control unit 31 acquires the teacher data 4a by accepting a user's request to classify certain explanatory variable data 40 into the natural payment data 41 . The control unit 31 acquires the teacher data 4b by accepting from the user that certain explanatory variable data 40 is classified into the telephone demanding payment data 42 . The control unit 31 acquires the teacher data 4c by accepting from the user that certain explanatory variable data 40 is classified as unpaid telephone reminder data 43 . The control unit 31 acquires a plurality of teacher data 4a, 4b, and 4c by classifying the plurality of explanatory variable data 40 into the explanatory variable data 40 described above. Note that the classification of the explanatory variable data 40 may be performed by the server device 2 .

A learning model 5 is a learning model used for machine learning. The learning model 5 is generated/updated by performing learning by any machine learning method. Any machine learning technique can solve the classification problem, such as logistic regression, decision trees, random forests, gradient boosting decision trees, neural networks, and the like.

In the first embodiment, the control unit 21 causes the learning model 5 to learn by an arbitrary machine learning method using the explanatory variable data 40 as the explanatory variable and the telephone demanding payment data 42 as the objective variable. When the learning ends, the control unit 21 acquires the learned learning model 7 . From a certain point of view, the control unit 21 determines whether the explanatory variable data 40 for each of the plurality of claims and whether the payment was made or the dunning action was executed as a result of executing the dunning action to the obligor of each of the plurality of claims. A learned learning model 7 is obtained by learning objective variable data indicating whether or not payment was made as a result of not doing so.

Next, explanatory variable data 40, which is an explanatory variable in machine learning, will be described with reference to FIG. FIG. 8 is a diagram showing an example of explanatory variable data 40 according to the first embodiment. Each data related to the explanatory variable data 40 is used as an explanatory variable correlated with class A, class B and class C. In the following description, it is assumed that the control unit 21 acquires each data related to the explanatory variable data 40, but the control unit 31 may acquire the data in place of the control unit 21. FIG.

The explanatory variable data 40 is data relating to bills to debtors. The explanatory variable data 40 includes data that can be specified using any method using data that can be obtained from the invoice. Arbitrary methods include a method of referring to a database, a method of acquiring by scraping, a method of acquiring by user input, and the like. The explanatory variable data 40 may include debtor data 400, seller data 401, billing amount data 402, agent data 403, billing date data 404, payment deadline data 405, and transaction object data 406. good.

Debtor data 400 and seller data 401 will be described later. The billed amount data 402 is billed amount data, such as "39,500 yen". The agent data 403 is data for identifying the agent Z, and includes data such as the agent Z's name and address. The billing date data 404 is data relating to the billing date, such as "October 30, 2020". The payment deadline data 405 is data relating to the payment deadline, and is data such as “November 30, 2020”, for example. The transaction target data 406 is data on the content of the product or service that is the target of the transaction, such as "product A".

The obligor data 400 is data associated with the obligor Y, and includes at least one of data on the attributes of the obligor Y and data on the performance of the obligor Y's transactions. Here, the data relating to the transaction record of the debtor Y is abbreviated as debtor record data. Data that can be acquired in association with the debtor data 400 is included in the debtor data 400 .

The attribute data of the obligor Y is acquired by, for example, the following method. The control unit 21 acquires the data of the name of the debtor Y and the data of the address of the debtor Y from the debtor data 400 . The control unit 21 identifies the debtor Y based on the data of the name of the debtor Y and the data of the address of the debtor Y. If the specified obligor Y is a corporation, the control unit 21 uses any of the above-described methods to determine whether the obligor Y's capital, number of employees, industry, elapsed time since establishment, listing status, corporate number etc. is acquired as attribute data of the obligor Y. For example, as an attribute of the obligor of the target receivables, if the obligor is a listed company, it is highly likely that the obligor will make payments similar to those of other listed companies included in the past data. can think.

The control unit 21 identifies based on at least one of the debtor data 400, billing amount data 402, agent data 403, billing date data 404, payment deadline data 405, and transaction object data 406. For debtor Y, obtain any debtor performance data from past processed claims linked to debtor Y. For example, when acquiring the debtor performance data related to the amount, the control unit 21 acquires the debtor performance data related to the amount based on the debtor data 400 and the billed amount data 402 . For example, when acquiring the debtor performance data related to the agent Z, the control unit 21 acquires the debtor performance data related to the agent Z based on the debtor data 400 and the agent data 403. . For example, when acquiring the debtor performance data related to the period, the control unit 21 acquires the debtor performance data related to the period based on the debtor data 400 and the billing date data 404 or the payment deadline data 405. do. For example, when acquiring the obligor track record data related to the transaction object, the control unit 21 acquires the obligor track record data related to the transaction object based on the obligor data 400 and the transaction object data 406 .

Obligor performance data includes, for example, data related to the number of days between the date of billing and the due date of payment, and how long the payment was delayed from the due date for receivables processed in the past linked to obligor Y. and data on how long the payment was delayed after a reminder was issued.
Obligor performance data includes, for example, the number of claims processed in the past related to obligor Y, the number of claims paid without demand, the number of claims paid with demand, and the number of claims paid despite demand. It may include at least one data with the number of receivables not cleared (the number of non-payments). For example, it is generally believed that an obligor with a track record of on-time payments in the past will likely continue to make on-time payments in the future. In addition, obligors who have past records of making payments past the due date without demanding payment, and obligors who have made many payments as a result of demanding payment, are likely to make similar payments in the future. it is conceivable that.
The obligor performance data may include, for example, data on the details of the goods or services that were the subject of the transaction with respect to the past processed receivables associated with the obligor Y.
Obligor performance data includes, for example, data on the total number of reminders issued to obligor Y himself, and data on the total number of reminders issued for each claim processed in the past linked to obligor Y. may contain.
The obligor performance data may include, for example, data on whether or not a claim processed in the past associated with the obligor Y was paid as a result of dunning.
The obligor performance data is, for example, data on whether invoices for claims processed in the past associated with obligor Y were sent to obligor Y by e-mail or by mail. may contain. For example, the debtor Y who sent the invoice by mail has the actual invoice, so it is relatively easy to recognize the existence of the invoice. Nonetheless, debtor Y who sent the bill by mail may not have made payment, and thus may be more likely to be classified in class C.
Obligor performance data may include, for example, data about previously processed receivables associated with obligor Y within the most recent period of time. The most recent fixed period may be within one year, within one month, within one week, etc. from the time of forecasting the target claims.

The seller data 401 is data associated with the seller X, and includes attribute data of the seller X for each of the plurality of claims to be learned, and transaction performance data regarding how the claims of the seller X were paid. and/or Here, transaction performance data relating to how seller X's receivables have been paid is abbreviated as seller performance data. Data that can be acquired in association with the seller data 401 is included in the seller data 401 .

The attribute data of the seller X for each of the plurality of bonds to be learned is acquired, for example, by the following method. The control unit 21 acquires the name data of the seller X and the address data of the seller X from the seller data 401 . The control unit 21 identifies the seller X based on the seller X name data and the seller X address data. If the specified seller X is a corporation, the control unit 21 uses any of the above-described methods to determine whether or not the seller X has capital, number of employees, industry, elapsed time since establishment, listing status, corporate number, etc. Data is acquired as attribute data of obligor Y.

The control unit 21 identifies the seller based on at least one of the seller data 401, billing amount data 402, agent data 403, billing date data 404, payment deadline data 405, and transaction object data 406. Regarding X, arbitrary seller performance data is acquired from past processed credits linked to seller X. For example, when obtaining the actual seller data related to the amount, the control unit 21 acquires the actual seller data related to the amount based on the seller data 401 and the billed amount data 402 . For example, when acquiring seller performance data related to agent Z, control unit 21 acquires seller performance data related to agent Z based on seller data 401 and agent data 403 . For example, when obtaining period-related seller performance data, the control unit 21 acquires period-related seller performance data based on the seller data 401 and the billing date data 404 or payment deadline data 405 . For example, when obtaining seller performance data related to the transaction object, the control unit 21 acquires the seller performance data related to the transaction object based on the seller data 401 and the transaction object data 406 .

The seller performance data includes, for example, data regarding the number of days between the billing date and the payment due date, and data regarding how long the payment was delayed from the payment due date for claims processed in the past associated with seller X. and data on how long there was a delay between making a reminder and getting paid.
The seller performance data includes, for example, the number of claims processed in the past related to seller X, the number of claims paid without demand, the number of claims paid with demand, and the number of claims not paid despite demand. and the number of receivables received (the number of unpaid claims). For example, obligors associated with dunned but not paid receivables are highly correlated with Class C.
The seller performance data may include, for example, data on the details of the goods or services that were the subject of transactions with respect to claims processed in the past that are tied to seller X.
The seller performance data includes, for example, data on the total number of reminders issued to the debtor Y linked to the seller X with respect to the claims processed in the past linked to the seller X, It may also include data on how many reminders have been issued in total for each bond.
The seller performance data may include, for example, data as to whether or not a claim processed in the past associated with the seller X was paid as a result of dunning.
The seller performance data may include, for example, data on the average amount of sales of seller X for receivables tied to seller X and processed in the past.
The seller performance data may include, for example, data on trends of debtors associated with seller X with respect to previously processed receivables associated with seller X. The obligor trend data is, for example, data indicating that many obligors who have transactions with the seller X include obligors in the restaurant business.
The seller performance data may include, for example, the performance data of seller X with respect to agent Z with respect to past processed receivables associated with seller X. The performance data of seller X with respect to agent Z includes, for example, data regarding how much sales were made to agent Z as a result of agent Z representing seller X. FIG.
The seller performance data may include, for example, data on past processed receivables tied to seller X within the most recent fixed period of time. The most recent fixed period may be within one year, within one month, within one week, etc. from the time of forecasting the target claims.

The explanatory variable data 40 includes data on the attributes of the debtor Y, data on the performance of transactions of the debtor Y, and data on the seller X for each of the plurality of debts to be learned at the time of examination of the debt in factoring. It may also include attribute data and transaction performance data regarding how seller X's receivables have been paid. In addition, in the examination of receivables in factoring, data on attributes of debtor Y, data on transaction performance of debtor Y, data on attributes of seller X for each of the plurality of receivables to be learned, and data on attributes of seller X Claims are reviewed based on transaction performance data and how the claims have been paid. The explanatory variable data 40 may include the results of examination of this claim. In addition, for the explanatory variable data 40 other than the data at the time of examination in factoring, data that can be acquired based on the time of execution of learning can basically be used, but it is not limited to this.

Next, with reference to FIG. 9, the details of the process of inputting the target bond data 6 of the activity A10 of FIG. 6 into the learned learning model 7 and outputting the output data 8 will be described. FIG. 9 is a diagram showing an example of a process of making predictions using the learned learning model 7 according to the first embodiment.

FIG. 9 includes target bond data 6, a learned learning model 7, and output data 8. FIG. The target bond data 6 is data relating to a bond for which it is predicted whether or not payment to the bond will be made. The learned learning model 7 is based on the explanatory variable data 40 for each of the plurality of claims and whether or not payment was made as a result of executing the dunning action to the obligor of each of the plurality of claims. It is a learning model obtained by learning the objective variable data indicating whether the payment was made as a result of the payment. The output data 8 is data indicating the probability that each obligor will make a payment to a plurality of prediction target claims when a dunning action is executed by telephone to each obligor of a plurality of prediction target claims. be. The output data 8 may be a telephone demanding payment data table T1 in tabular form. The telephone demanding payment data table T1 will be described in detail with reference to FIG.

The control unit 21 outputs output data 8 by inputting a plurality of target bond data 6 to a learned learning model 7 . According to a certain point of view, the control unit 21 inputs a plurality of target bond data 6 to a learned learning model 7, and executes a dunning action by telephone to each obligor of a plurality of prediction target bonds. output the probabilities that each obligor will make payments to the multiple bonds to be predicted. In addition, when outputting the probability of demanding actions such as e-mails and visits in addition to telephone calls, the control unit 21 inputs a plurality of target bond data 6 to the learned learning model 7 to generate a plurality of prediction targets. output the probabilities that each obligor will make payment to a plurality of prediction target obligors when a dunning action is executed for each obligor of each obligor.

Also, the control unit 21 may output the output data 8 at a predetermined timing. The predetermined timing may be every day, the beginning of the month, every Monday, or the like, and is a timing that the user can arbitrarily set. From a certain point of view, the control unit 21 inputs the target bond data 6 to the learned learning model 7 at a predetermined timing, so that the obligor of the target bond to be predicted performs a dunning action or a dunning action. A prediction is made as to whether the predicted claim will be paid by the debtor if the action is not performed. Further, the control unit 21 makes predictions at regular time intervals. At this time, the claims to be predicted at the timing of the prediction are the claims whose payment due dates have passed.

Next, using FIG. 10, details of the telephone demand payment data table T1 output in activity A13 of FIG. 6 will be described. FIG. 10 is a diagram showing an example of the telephone demanding payment data table T1 according to the first embodiment.

The telephone demanding payment data table T1 is data in a table format showing the data of the claim and the probability of whether the obligor of the claim falls under class B or not. The telephone dunning payment data table T1 is referred to by the user associated with agent Z. The user using the client device 3 refers to the telephone demanding payment data table T1 and performs the demanding work. The telephone demand payment data table T1 includes bond ID data T10, seller data T11, debtor data T12, debt amount data T13, unpaid amount data T14, payment date data T15, payment deadline data T16, and class B probability data T17 are included.

The bond ID data T10 is data of an ID specifying a bond, such as "100001". The seller data T11 is data of the name of the legal entity or natural person of the seller X, such as "Ringo Sales Co., Ltd.". The obligor data T12 is data of the name of the legal entity or natural person of the obligor Y, such as "Zaibatsu Co., Ltd.". The debt amount data T13 is data of the amount of money to be paid for the claim, such as "264,000 yen" and "1,465,000 yen".

The unpaid amount data T14 is data of an unpaid amount for a claim, such as "264,000 yen". "0" is given to the unpaid amount data T14 relating to the claim for which payment has been made. The remaining unpaid amount may be given to the outstanding amount data T14 relating to partially paid claims. For example, it is assumed that only 700,000 yen has been paid for a bond whose bond ID data T10 is "100003". In this case, the debt amount data T13 is "1,465,000 yen" and the unpaid amount data T14 is "765,000 yen". The payment date data T15 is data of the payment date, such as "2022/12/15". "None" is given to the payment date data T15 relating to unpaid claims. The payment deadline data T16 is data of payment deadline, such as "2022/11/30". Class B probability data T17 is data indicating the probability that the obligor is class B.

According to the above, it is possible to provide a technique for making reminders more efficient using machine learning. In addition, human resources can be efficiently distributed to claims that are highly likely to be paid by dunning.

Next, Embodiments 2 to 4 will be described. In Embodiments 2 to 4, other examples of predicting class A, class B, and class C using various target variable data will be described. From a certain point of view, the control unit 21 acquires a plurality of target bond data 6, which are data on a plurality of prediction target bonds. By inputting the plurality of target bond data 6 to the learned learning model 7, the control unit 21 performs a dunning action on each obligor of the plurality of prediction target bonds as shown in FIGS. 11 to 13. output the probabilities that each obligor will make payment to a plurality of predicted claims if the demanding action is executed or if the dunning action is not executed.

[Embodiment 2]
In the second embodiment, a case will be described where it is predicted whether or not the claim is a claim (class C) that will not be paid even if demanding is made. The system configuration of the second embodiment is similar to that of the first embodiment. The hardware configurations of the server device and the client device of the second embodiment are the same as the hardware configurations of the server device 2 and the client device 3 of the first embodiment. Information processing in the second embodiment is substantially the same as in the first embodiment, and differences from the first embodiment will be described below.

In the second embodiment, the control unit 21 uses the explanatory variable data 40 as the explanatory variable and the unpaid telephone reminder data 43 as the objective variable to make the learning model 5 learn by any machine learning technique.

The control unit 21 of the second embodiment inputs a plurality of target bond data 6 to the learned learning model 7 to output the output data 8 including the telephone reminder unpaid data table T2 shown in FIG. From a certain point of view, the control unit 21 inputs a plurality of target bond data 6 to a learned learning model 7 to execute a demanding action by telephone to each obligor of a plurality of prediction target bonds. output the probabilities that each obligor does not make payments to the multiple bonds to be predicted. In addition, when outputting the probability of demanding actions such as e-mails and visits in addition to telephone calls, the control unit 21 inputs a plurality of target bond data 6 to the learned learning model 7 to generate a plurality of prediction targets. output the probabilities that each obligor does not pay the prediction target multiple claims even if the demanding action is executed for each obligor of each of the claims.

FIG. 11 is a diagram showing an example of the telephone reminder unpaid data table T2 according to the second embodiment. The unpaid telephone demand data table T2 is data in a tabular form showing data of claims and probabilities of whether the obligors of the claims belong to class C or not. Unlike the telephone demand payment data table T1, the telephone demand unpaid data table T2 does not have class B probability data T17, but has class C probability data T20. Class C probability data T20 is data indicating the probability that the obligor is class C.

As a result, it is possible to promptly take special measures for a claim that is highly likely not to be paid even if it is demanded, so that the probability of payment to the claim can be increased.

[Embodiment 3]
In the third embodiment, a case of predicting whether or not the claim is overdue but will be paid naturally (class A) will be described. The system configuration of the third embodiment is similar to that of the first embodiment. The hardware configurations of the server device and the client device of the third embodiment are the same as the three hardware configurations of the server device 2 and the client device of the first embodiment. Information processing in the third embodiment is substantially the same as in the first embodiment, and differences from the first embodiment will be described below.

In the third embodiment, the control unit 21 causes the learning model 5 to learn by any machine learning method using the explanatory variable data 40 as the explanatory variable and the natural payment data 41 as the objective variable.

The control unit 21 of Embodiment 3 inputs a plurality of target bond data 6 to the learned learning model 7, thereby outputting the output data 8 including the natural payment data table T3 shown in FIG. From a certain point of view, the control unit 21 inputs a plurality of target bond data 6 to a learned learning model 7 to execute a demanding action by telephone to each obligor of a plurality of prediction target bonds. Output the probabilities that a plurality of predicted bonds will be paid by each obligor, even if they do not exist. In addition, when outputting the probability of demanding actions such as e-mails and visits in addition to telephone calls, the control unit 21 inputs a plurality of target bond data 6 to the learned learning model 7 to generate a plurality of prediction targets. outputs the probabilities that each debtor will pay the predicted multiple debts even if the obligors of each debtor do not perform a dunning action.

FIG. 12 is a diagram showing an example of the natural payment data table T3 according to the third embodiment. The natural payment data table T3 is data in the form of a table showing the data of the bond and the probability of whether the bond falls under class A or not. Natural payment data table T3 does not have class B probability data T17, but has class A probability data T30, as compared with telephone demand payment data table T1. The class A probability data T30 is data indicating the probability that the claim is class A.

As a result, it is possible to take the option of not issuing a demand for a claim that is highly likely to be paid even without a demand, thereby making efficient use of human resources.

In Embodiments 1 to 3 described above, the control unit 21 may output the result of determination as to whether the class is A to C based on the respective probabilities of classes A to C. FIG. For example, in the third embodiment, when the probability of class A is 50% or more, the control unit 21 outputs data such as "predicted as class A", and when the probability of class A is less than 50%, the control unit 21 outputs " Data such as "predicted not to be class A" may be output.

[Embodiment 4]
In the fourth embodiment, a case of predicting which of class A, class B, and class C the bond belongs to will be described. The system configuration of the fourth embodiment is similar to that of the first embodiment. The hardware configurations of the server device and the client device of the fourth embodiment are the same as the hardware configurations of the server device 2 and the client device 3 of the first embodiment. Information processing in the fourth embodiment is substantially the same as in the first embodiment, and differences from the first embodiment will be described below.

In the third embodiment, the control unit 21 uses the explanatory variable data 40 as an explanatory variable, the natural payment data 41, the telephone demand payment data 42, and the telephone demand unpaid data 43 as objective variables, and generates a learning model using an arbitrary machine learning technique. Let 5 learn.

The control unit 21 of the fourth embodiment inputs a plurality of target bond data 6 to the learned learning model 7, thereby outputting the output data 8 including the bond prediction data table T4 shown in FIG. The bond prediction data table T4 is data in a table format showing data on bonds and probabilities to which class A, class B, and class C the bond belongs. According to one aspect, the control unit 21 inputs a plurality of target bond data 6 to the learned learning model 7 . The control unit 21 determines the probability that each of the obligors of the plurality of prediction target claims will be paid by each of the obligors when the obligors of the plurality of prediction target credits are demanded by telephone. is the probability that each obligor does not pay the predicted receivables even if the obligor of each of the receivables is dunned by telephone, and the probability that each of the predicted receivables and the probability that each debtor will make payment to a plurality of prediction target claims even if the demanding action is not executed by telephone to each debtor.

FIG. 13 is a diagram showing an example of a bond prediction data table T4 according to the fourth embodiment. Compared to the telephone demand payment data table T1, the natural payment data table T3 does not have class B probability data T17, but instead has class A relative probability data T40, class B relative probability data T41, and class B relative probability data T41. C relative probability data T42.

The class A relative probability data T40 is data indicating the relative probability that the bond is class A. The class B relative probability data T41 is data indicating the relative probability that the bond is class B. The class C relative probability data T42 is data indicating the relative probability that the bond is class C. Since each of the relative probability data T40, T41, T42 is relative, the sum of the respective probabilities of class A, class B, and class C for a certain bond is 100%. For example, for a bond with a bond ID of “100001”, the class A relative probability data T40 is “0.30”, the class B relative probability data T41 is “0.50”, and the class C relative probability data T42 is Adding "0.20" results in "1".

As a result, it is possible to take the option of not issuing a demand for a claim that is highly likely to be paid even without a demand, thereby making efficient use of human resources. In addition, human resources can be efficiently distributed to claims that are highly likely to be paid by dunning. Furthermore, it is possible to promptly take special measures for claims that are highly likely not to be paid even if demanded, thereby increasing the probability of payment to the claims.

The lists as shown in FIGS. 10 to 13 are output on "2022/12/1" after the payment deadline "2022/11/30". By doing so, it is possible to prevent the data on bills that have already been paid from being listed. That is, the plurality of prediction target claims may be only past due dates. In addition, lists such as those shown in FIGS. 10 to 13 can be rearranged in descending or descending order of probability. Lists such as those shown in FIGS. 10 to 13 are displayed, for example, on a screen used by an operator who performs demanding work by telephone, and the operator can examine the priority of demanding work by telephone while viewing the list. can.

It should be noted that the probabilities in FIGS. 10-13 are illustrative numbers and not the actual values obtained by the model of this disclosure. Note that data other than the probabilities shown in FIGS. 10-13 are also exemplary data.

[Embodiment 5]
In the fifth embodiment, a case where one computer executes all information processing will be described.

Unlike the first embodiment, the information processing system of the fifth embodiment does not have a server device and a network, and is composed only of client devices. According to one aspect, a system exemplified by an information processing system is comprised of one or more devices or components. Therefore, a server device or a client device alone is included in a system exemplified as an information processing system. The hardware configuration of the client device of the fifth embodiment is the same as the configuration of the client device 3 of the first embodiment.

In the fifth embodiment, unlike the first embodiment, information is not transmitted and received between the server device and the client device, and all processing is performed by the client device. For example, the information processing system of the fifth embodiment does not execute information transmission/reception processing in A2-A3, A5-A6, A8-A9, and A11-A12 in FIG. The client device of the fifth embodiment executes information processing other than the information transmission/reception processing shown in FIG.

As a result, it is possible to provide a technique for making reminders more efficient using machine learning without connecting to a network.

[Embodiment 6]
In the sixth embodiment, a case will be described in which a client device executes either information processing such as acquisition of a learned learning model 7 or prediction using a learned learning model 7 . The system configuration of the sixth embodiment is similar to that of the first embodiment. The hardware configurations of the server device and the client device of the sixth embodiment are the same as the hardware configurations of the server device and the client device of the first embodiment.

In the sixth embodiment, the processing related to acquisition of the learned learning model 7 of A1 to A6 may be executed only by the client device. In other words, the information processing system of the sixth embodiment does not execute the information transmission/reception processing in A2 to A3 and A5 to A6 in FIG. The client device of the sixth embodiment executes information processing other than the information transmission/reception processing in A1 to A6 in FIG. In this case, the control unit 31 uploads the learned learning model 7 to the server device 2 .

Further, in the sixth embodiment, the processing related to prediction using the learned learning model 7 of A7 to A13 may be executed only by the client device. In other words, the information processing system of the sixth embodiment does not execute the information transmission/reception processes in A8 to A9 and A11 to A12 in FIG. The client device of the sixth embodiment executes information processing other than the information transmission/reception processing of A7 to A13 in FIG. In this case, the control unit 31 downloads the learned learning model 7 from the server device 2 .

Accordingly, information processing can be executed in various forms of information processing systems.

[others]
In Figure 2, receivables that are not paid within the due date are receivables that are paid naturally even though the due date has passed (Class A), receivables that are paid after demanding (Class B), and receivables that are not paid even after demanding. Although it has been described as being classified into any one of the claims (class C), it may be further classified into class D, which is not classified into any of class A, class B, or class C. Moreover, when it is classified into class D, it may be machine-learned. Note that class D is, for example, a claim that has not been demanded or deposited.

The objective variable data includes data regarding whether the claim falls under class D or not. By inputting a plurality of target bond data 6 to a learned learning model 7, the control unit 21 outputs the probability of whether each of the plurality of prediction target bonds falls under class D.

Also, in the embodiment, the probabilities of classes A to C are output, but in a modified example, the predicted payment date may be output. In that case, the control unit 21 acquires a trained model by using the interval between the payment deadline and the actual payment date as the objective variable. In that case, the control unit 21 may use, in addition to the explanatory variables described in the first embodiment, data indicating which of the classes A to C the object is classified into as an explanatory variable.

Further, in the embodiment, the claim is explained as a bill, but it may be in the form of a contract, a bill, an invoice, a slip, etc. as long as payment is requested.

Regarding the information processing system 1 according to the above-described embodiment, it may be a program that causes a computer to function as the control unit 31 of the information processing system 1 . Alternatively, it may be an information processing method executed by the information processing system 1 .

Furthermore, it may be provided in each aspect described below.

(1) In an information processing system, using a learning model, whether the debtor pays the debt when a dunning action is performed or when the dunning action is not performed by the debtor The learning model includes explanatory variable data for each of a plurality of claims and whether or not a payment was made as a result of executing a dunning action to the obligor of each of the plurality of claims. a learning model obtained by learning objective variable data representing whether or not payment has been made as a result of not executing a dunning action, and the result of whether or not payment has been made; By acquiring the target credit data and inputting the target credit data into the learning model, the obligor of the prediction target credit is informed whether or not to perform the dunning action. an information processing system that predicts whether the claim to be predicted will be paid by

According to such a configuration, it is possible to provide a technique for making reminders more efficient using machine learning.

(2) In the information processing system described in (1) above, the control unit acquires a plurality of target bond data, which are data about a plurality of prediction target bonds, and applies the plurality of target bond data to the learning model. is input to the prediction target obligors, if the obligors of the plurality of prediction target credits are to perform the dunning action or if the dunning action is not performed, the respective obligors to the prediction target credits. An information processing system that outputs each probability that a payment will be made.

According to such a configuration, it is possible to provide a technique for making demanding more efficient by using machine learning for a plurality of claims.

(3) In the information processing system described in (2) above, the objective variable data is stored in the claim by the debtor when the debtor makes a demanding call after the payment deadline has passed. and the control unit inputs the plurality of target receivables data into the learning model to provide the obligors of each of the plurality of receivables to be predicted. An information processing system for outputting respective probabilities that the plurality of claims to be predicted will be paid by each of the obligors when executing a dunning action by telephone.

According to such a configuration, it is possible to efficiently distribute human resources to claims that are highly likely to be paid by dunning.

(4) In the information processing system according to any one of the above (2) to (3), the objective variable data is when the debtor executes a dunning action by telephone after the payment deadline has passed. includes unpaid telephone demand data regarding whether or not the debtor has paid the debt, and the control unit inputs the plurality of target debt data to the learning model to obtain the prediction target An information processing system for outputting respective probabilities that the respective obligors will not make payments on the plurality of claims to be predicted even if the obligors of the plurality of claims are demanded by telephone.

With such a structure, it is possible to take special measures immediately for claims that are highly likely not to be paid even if demand is made, thereby increasing the probability of payment to the claims. can.

(5) In the information processing system according to any one of the above (2) to (4), the objective variable data indicates that the debtor did not perform a dunning action by telephone after the payment deadline has passed. including natural payment data relating to whether the debtor has paid the claim in the event of a failure, and the control unit inputs the plurality of target claim data into the learning model to determine the plurality of prediction target An information processing system for outputting probabilities that the respective obligors will pay the plurality of claims to be predicted without executing a demanding action by telephone to the respective obligors of the claims.

According to such a configuration, it is possible to take the option of not demanding payment for a claim that is likely to be paid without demanding, and to efficiently utilize human resources.

(6) In the information processing system according to any one of the above (2) to (5), the objective variable data is when the debtor executes a dunning action by telephone after the payment deadline has passed. telephone dunning payment data regarding whether payment has been made on said receivable by said debtor; telephone dunning outstanding data as to whether payment to the debtor has failed to be made and payment to the receivable by the debtor in the event of failure to perform a telephone dunning action against the debtor after the payment is overdue; and the control unit inputting the plurality of target credit data into the learning model so as to demand the obligors of the plurality of prediction target credits by telephone. Probability of payment to the plurality of predicted claims by each of the obligors when the action is executed, and execution of a dunning action by telephone to each of the obligors of the plurality of predicted claims. the probability that the respective obligors will not make payments on the plurality of predicted claims even if the respective obligors do not make payments, and the probability that the respective obligors of the plurality of predicted claims will not be paid even if the demanding action is not performed by telephone. an information processing system for outputting a probability that each of the obligors will pay the plurality of claims to be predicted.

According to such a configuration, it is possible to take the option of not demanding payment for a claim that is likely to be paid without demanding, and to efficiently utilize human resources. In addition, human resources can be efficiently distributed to claims that are highly likely to be paid by dunning. Furthermore, it is possible to promptly take special measures for claims that are highly likely not to be paid even if demanded, thereby increasing the probability of payment to the claims.

(7) In the information processing system according to any one of (1) to (6) above, the explanatory variable data is data relating to an invoice to the debtor, and includes data of the invoiced amount and data of the transaction. An information processing system containing data on the content of goods or services to be covered and data on payment deadlines.

According to such a configuration, it is possible to predict with a high degree of accuracy what kind of obligor the obligor is, since the bill-related data is used.

(8) In the information processing system according to any one of (1) to (7) above, the explanatory variable data includes seller attribute data for each of the plurality of claims to be learned, and/or transaction performance data relating to how the receivables of the Company were paid.

According to such a configuration, the type of debtor can be predicted with high accuracy because the data on the number of unpaid debts is used.

(9) In the information processing system according to any one of (1) to (8) above, the explanatory variable data includes data on the attributes of the debtor and data on the transaction performance of the debtor. An information processing system including at least one.

According to such a configuration, it is possible to predict with a high degree of accuracy what kind of obligor the borrower is, since the data on the examination is used.

(10) In the information processing system according to any one of (1) to (9) above, the control unit performs the prediction at regular time intervals, and the prediction target at the timing when the prediction is performed An information processing system in which receivables are receivables that are past due for payment.

According to such a configuration, the user can use the prediction result at a predetermined timing.

(11) An information processing system comprising a control unit, wherein the control unit performs a dunning action on explanatory variable data on each of a plurality of claims and on each of the obligors of the plurality of claims. An information processing system that acquires a learning model by learning objective variable data indicating whether payment was made as a result of payment or payment was made as a result of not performing a dunning action.

According to such a configuration, it is possible to provide a technique for making reminders more efficient using machine learning.

(12) A program for causing a computer to function as the control section of the information processing system according to any one of (1) to (11) above.

(13) An information processing method executed by an information processing system, comprising each process executed by the control unit of the information processing system according to any one of (1) to (11) above. .
Of course, this is not the only case.

Finally, while various embodiments of the invention have been described, these have been presented by way of example and are not intended to limit the scope of the invention. The novel embodiment can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. The embodiment and its modifications are included in the scope and gist of the invention, and are included in the scope of the invention described in the claims and equivalents thereof.

1: information processing system 2: server device 21: control unit 22: storage unit 23: communication unit 3: client device 31: control unit 32: storage unit 33: communication unit 34: input unit 35: output unit 4a: teacher data 4b : Teacher data 4c : Teacher data 40 : Explanatory variable data 41 : Natural payment data 42 : Telephone reminder payment data 43 : Telephone reminder unpaid data 400 : Debtor data 401 : Seller data 402 : Billed amount data 403 : Agent data 404 : Billing date data 405 : Payment deadline data 406 : Transaction object data 5 : Learning model 6 : Object bond data 7 : Learning model 8 : Output data I : Billing means M1 : Price M2 : Price N : Notification means NW : Network P : Product T1: Telephone reminder payment data table T10: Transaction ID data T11: Seller data T12: Debtor data T13: Debt amount data T14: Unpaid amount data T15: Payment date data T16: Payment deadline data T17: Probability data T2: Telephone reminder Unpaid data table T20: Probability data T3: Natural payment data table T30: Probability data T4: Debtor prediction data table T40: Relative probability data T41: Relative probability data T42: Relative probability data X: Seller Y: Debtor Z: Agent person

Claims (13)

An information processing system that uses a learning model to make a prediction as to whether the debtor will pay the debt if the debtor of the debt is dunned or not dunned. having a control unit,
The learning model includes explanatory variable data for each of a plurality of receivables and whether payment was made as a result of performing a dunning action or not as a result of not performing a dunning action for each obligor of the plurality of receivables. A learning model obtained by learning objective variable data representing whether or not payment has been made, and
The control unit
Acquire target receivables data for the receivables to be predicted,
By inputting the target credit data into the learning model, when the obligor of the prediction target credit performs a dunning action or does not perform the dunning action, make predictions about whether payments for
Information processing system. In the information processing system according to claim 1,
The control unit
Acquiring a plurality of target receivables data, which is data about a plurality of prediction target receivables,
By inputting the plurality of target receivables data into the learning model, each of the obligors of the plurality of receivables to be predicted performs a dunning action or does not perform a dunning action. respectively outputting the probabilities that the plurality of claims to be predicted will be paid by
Information processing system. In the information processing system according to claim 2,
The objective variable data includes telephone demand payment data relating to whether the debtor has paid the claim when a demanding action is performed by telephone against the debtor after the payment deadline has passed,
The control unit
By inputting the plurality of target receivable data into the learning model, when the obligor of each of the plurality of receivables to be predicted performs a dunning action by telephone, Output the probabilities of payments to multiple bonds, respectively,
Information processing system. In the information processing system according to claim 2,
The objective variable data includes telephone demand unpaid data regarding whether or not the debtor has not paid the claim when the demanding action is executed by telephone to the debtor after the payment deadline has passed. ,
The control unit
By inputting the plurality of target credit data into the learning model, even if the obligors of the plurality of prediction target credits are demanded by telephone, Output the probabilities of non-payment of multiple bonds,
Information processing system. In the information processing system according to claim 2,
The objective variable data includes natural payment data relating to whether the receivable has been paid by the obligor if the obligor did not perform a telephone reminder action after the payment deadline,
The control unit
By inputting the plurality of target credit data into the learning model, each of the prediction target credit data is obtained by each of the prediction target debtors without executing a demanding action by telephone to each debtor of the plurality of prediction target credits. outputting respective probabilities of payment to the plurality of bonds;
Information processing system. In the information processing system according to claim 2,
The objective variable data is
telephone dunning payment data relating to whether payment of said claim has been made by said debtor when performing a telephone dunning action against said debtor after the payment is overdue;
unpaid telephone reminder data regarding whether or not the debtor did not pay the claim when the debtor performed a dunning action by telephone after the payment deadline;
natural payment data relating to whether payment of said receivables was made by said debtor in the absence of a telephone reminder to the debtor after payment was due;
The control unit
By inputting the plurality of target bond data into the learning model,
a probability that each of the obligors of the plurality of prediction target claims will be paid by each of the obligors when the demanding action is performed by telephone to each obligor of the plurality of prediction target claims;
a probability that the respective obligors of the plurality of prediction target claims will not be paid even if the obligors of the plurality of prediction target claims are demanding by telephone;
and a probability that each of the obligors of the plurality of prediction target claims will be paid even if the obligors of the plurality of prediction target claims are not demanded by telephone. do,
Information processing system. In the information processing system according to claim 1,
The explanatory variable data is data related to an invoice to the debtor, and includes data on the invoiced amount, data on the details of the goods or services to be traded, and data on the payment deadline.
Information processing system. In the information processing system according to claim 1,
The explanatory variable data includes at least one of seller attribute data for each of the plurality of receivables to be learned and transaction performance data regarding how the seller's receivables were paid,
Information processing system. In the information processing system according to claim 1,
The explanatory variable data includes at least one of data on the attributes of the debtor and data on the performance of transactions of the debtor,
Information processing system. In the information processing system according to claim 1,
The control unit
making the prediction at regular time intervals;
The claim to be predicted at the timing of the prediction is a claim whose payment due date has passed.
Information processing system. An information processing system,
having a control unit,
The control unit
Explanatory variable data for each of a plurality of claims and whether payment was made as a result of performing a dunning action or not as a result of not performing a dunning action to each obligor of the plurality of claims. Acquire a learning model by learning target variable data that indicates the result of
Information processing system. a program,
A program for causing a computer to function as the control unit of the information processing system according to any one of claims 1 to 11. An information processing method executed by an information processing system,
Each process executed by the control unit of the information processing system according to any one of claims 1 to 11,
Information processing methods.

Images