JP6797956B2 - Estimator, estimation method and program - Google Patents

Description

The present invention relates to an estimation device, an estimation method and a program. More specifically, the present invention accumulates or machine-learns image data generated based on the dynamic data of the accounts of various lenders associated with a plurality of business situation patterns, and the dynamics of the account of a specific lender. By analyzing image data based on data based on accumulated or machine-learned image data, an estimation device and estimation method for predicting the occurrence of large fluctuations in business conditions, especially default (default). And about the program.

Traditionally, the credit status of lenders has been regularly monitored. Banks change the management situation of lenders by acquiring financial numerical data from financial statements provided by lenders and by having corporate managers regularly conduct hearings with corporate managers. I'm judging if it's not.

Attempts have been made not only to make such judgments manually, but also to use computers. For example, a device for calculating the creditworthiness of a lender based on data on the financial status of the lender has been known (Patent Document 1).

In recent years, machine learning methods using artificial intelligence (AI) have come to be used in various fields. It is also becoming common to apply a technique called deep learning, which is one of the methods supporting AI, which is an evolution of a neural network, to image classification. Among the deep learning models, the convolutional neural network (CNN) is known to be a model widely used for image and moving image recognition.

JP-A-2017-016652

The financial numerical data used for monitoring the credit status of the lender could only be obtained once a year or once every six months, which is the timing when the financial statements are prepared. In addition, depending on the size of the company, financial statements may be submitted to the lender company after a considerable amount of time has passed since the closing month.

Since the business condition of a company changes rapidly due to external and internal factors, the credit status of the company often changes in a short period of time. Depending on external and / or internal factors, a company's credit status may change in a fairly short period of time, and in some cases may become a default (default) condition.

When evaluating the credit status of a lender based on numerical financial data, the lender company may not be able to perform the evaluation in a timely and appropriate manner due to the rapid changes in the actual business conditions. In such a case, the lender company could not properly support the lender for management improvement, and the default of the lender may make it difficult to collect the loan.

The present invention has been made to solve such a problem, and accumulates or machine-learns image data generated based on dynamic data of accounts of various lenders associated with a plurality of business situation patterns. By analyzing image data based on the dynamic data of the account of a specific lender based on accumulated image data or machine-learned image data, it is possible to detect significant changes in business conditions, especially whether or not defaults occur. It is an object of the present invention to provide an estimation device, an estimation method and a program for predicting.

The estimation device according to one aspect of the present invention reads the deposit / withdrawal detail data associated with the customer designated as the analysis target, and associates the variables included in the read deposit / withdrawal detail data with the first period. The first frequency distribution data and the second frequency distribution data associated with the second period are generated, and based on the first frequency distribution data and the second frequency distribution data, in the first period. The first image data including the associated first histogram and the second histogram associated with the second period is generated, and the first histogram and the second histogram included in the first image data are generated. The magnitude of the difference was identified, it was determined whether the magnitude of the identified difference was greater than or equal to a predetermined threshold, and it was determined that the magnitude of the difference was greater than or equal to the predetermined threshold. Under the condition that the first image data and the second image data are compared, the business situation of the customer is predicted.

The estimation method performed by the estimation device, which is another aspect of the present invention, reads the deposit / withdrawal detail data associated with the customer designated as the analysis target, and the variables included in the read deposit / withdrawal detail data. On the other hand, the first frequency distribution data associated with the first period and the second frequency distribution data associated with the second period are generated, and the first frequency distribution data and the second frequency are generated. To generate first image data including a first histogram associated with the first period and a second histogram associated with the second period based on the distribution data, and to generate the first image data. The magnitude of the difference between the first histogram and the second histogram included in the data is identified, and it is determined whether or not the magnitude of the identified difference is equal to or larger than a predetermined threshold value. By comparing the first image data with the second image data on the condition that the size of the data is determined to be equal to or larger than a predetermined threshold value, the business situation of the customer can be predicted. Including.

According to the present invention, financial numerical data has been used by analyzing image data based on dynamic data of a specific lender's account based on accumulated image data or machine-learned image data. It will be possible to predict large fluctuations in business conditions, especially the occurrence of defaults, at an earlier stage.

According to the present invention, since it becomes possible to predict the outlook of the business situation at an earlier stage than before, it becomes possible to provide the lender with timely and appropriate support for management improvement, and also. It will be possible to prevent the situation where it becomes difficult to collect the loan due to the default of the loan recipient.

A detailed understanding of the embodiments disclosed herein can be obtained from the following description illustrated in connection with the accompanying drawings.

It is a block diagram of the whole system including the estimation apparatus which concerns on embodiment of this invention. It is a system block diagram of the estimation apparatus which concerns on embodiment of this invention. It is a figure which shows an example of the data structure of the customer master which concerns on embodiment of this invention. It is a figure which shows an example of the data structure of the deposit / withdrawal details which concerns on embodiment of this invention. It is a figure which shows an example of the data structure of the learning numerical data which concerns on embodiment of this invention. It is a figure which shows an example of the data structure of the learning image data which concerns on embodiment of this invention. It is a figure explaining the image data generation processing and machine learning processing executed by an estimation apparatus. It is a figure explaining the prediction process executed by the estimation device. It is a figure which illustrates the process of generating the image data which includes the 1st histogram and the 2nd histogram.

(overall structure)
Hereinafter, embodiments of the present invention will be described in detail. FIG. 1 is a configuration diagram of the entire system including the estimation device 100 according to the embodiment of the present invention. The estimation device 100 is communicably connected to the banking system 110. As shown in FIG. 1, the estimation device 100 can be configured as an entity different from the banking system 110, or can be configured as an internal function of the banking system 110.

The banking system 110 is communicably connected to one or more customer terminals 120, one or more ATMs 130, and one or more other banking systems 140 via the Internet and known networks such as leased lines. .. In FIG. 1, only one customer terminal 120, one ATM 130, and one other bank system 140 are shown, but there may be a plurality of these.

In the present specification, the estimation device 100 is described as one device, but various processes executed by the estimation device 100 may be configured to be distributed and executed by a plurality of devices.

The estimation device 100 is associated with predetermined first period data and second period data for each of the variables included in the deposit / withdrawal statement data of various lender accounts associated with a plurality of business condition patterns. Generate and store each frequency distribution data. When there are a plurality of accounts associated with one lender, the estimation device 100 can generate frequency distribution data after performing name identification processing on them and collecting them into one.

The estimation device 100 uses the generated frequency distribution data to generate and store image data including a first histogram associated with the first period and a second histogram associated with the second period. The estimation device 100 can also make artificial intelligence (AI) machine-learn the generated image data.

In addition, the estimation device 100 has a predetermined first period for each of the variables included in the deposit / withdrawal detail data of the customer account of the target customer for predicting the occurrence of a drastic change in the business situation, particularly the occurrence of default. And generate each frequency distribution data associated with the data in the second period. When a plurality of customer accounts of the customer exist, the estimation device 100 can generate frequency distribution data after performing name identification processing on them and collecting them into one.

The estimation device 100 uses the generated frequency distribution data to generate image data including a first histogram associated with the first period and a second histogram associated with the second period. The estimation device 100 compares the generated image data with the stored image data, and determines in advance which business condition pattern (growth, normal, default) the customer to be predicted is likely to correspond to. Judgment is made according to the given criteria. Further, the estimation device 100 provides the generated image data to the AI, compares the generated image data with the machine-learned image data with the AI, and the customer to be predicted can correspond to which business situation pattern. It is possible to judge whether the property is high. The estimation device 100 can output these determination results.

The banking system 110 corresponds to a core banking system in a bank, and controls processing for customer accounts, exchange processing, ATM (Automated Teller's Machine) network, and connection processing with an external system. The bank system 110 provides a deposit / withdrawal statement of a designated customer account in response to an access from the customer terminal 120, and also responds based on a fund transfer (transfer, transfer) instruction from the customer terminal 120. Perform a transfer of funds from your account. The banking system 110 provides the balance information of the designated customer account in response to the access from the ATM 130, and also executes the transfer of funds from the corresponding customer account based on the fund transfer instruction from the ATM 130. The bank system 110 transmits a transfer request message to the other bank system 140 based on the transfer request from the customer, and also receives the transfer request message addressed to the customer account from the other bank system 140. The bank system 110 performs deposit / withdrawal processing on the customer account based on instructions from the customer terminal 120, the ATM 130, and the other bank system 140, and generates and stores the deposit / withdrawal detail data.

The customer terminal 120 is a customer-operated computer that has access to the banking system 110. The customer terminal 120 can be, for example, a computer having a communication function such as a personal computer (PC), a smartphone, or a tablet terminal, but is not limited to a specific device. The customer terminal 120 can access the bank system 110, receive the deposit / withdrawal statement data of the designated customer account, and transmit the fund transfer instruction from the customer account to the designated account to the bank system 110.

The ATM 130 can receive and display the balance of the specified customer account from the banking system 110 in response to the customer's operation, or can issue a fund transfer instruction generated based on the information input by the customer to the bank. It can be transmitted to the system 110.

The other bank system 140 is a banking system operated by another financial institution and includes a core banking system. The other bank system 140 receives a transfer request message from the bank system 110, performs a deposit process for the designated destination account, performs a withdrawal process from the designated account, and generates a transfer request message to generate a bank system. Send to 110.

(System configuration)
FIG. 2 is a system configuration diagram of the estimation device 100 according to the embodiment of the present invention. As shown in FIG. 2, the estimation device 100 has a control unit 201, a main storage unit 202, an auxiliary storage unit 203, an interface (IF) unit 204, and the like, like a general computer. The output unit 205 is provided. Further, the estimation device 100 includes a customer master 206, a deposit / withdrawal detail 207, a learning numerical data 208, and a learning image data 209 in the form of a file / database or the like. When the estimation device 100 and the banking system 110 are different entities, the customer master 206 and the deposit / withdrawal statement 207 may be received from the banking system 110 and processed synchronously.

The control unit 201, also called a central processing unit (CPU), controls each component of the estimation device 100 and calculates data, and also transmits various programs stored in the auxiliary storage unit 203 to the main storage unit 202. Read and execute. The main storage unit 202, also called a main memory, stores various received data, computer-executable instructions, and data after arithmetic processing by the instructions. The auxiliary storage unit 203 is a storage device typified by a hard disk (HDD) or the like, and is used when storing data or programs for a long period of time.

The embodiment of FIG. 2 describes an embodiment in which the control unit 201, the main storage unit 202, and the auxiliary storage unit 203 are provided inside the same computer, but as another embodiment, the estimation device 100 uses the control unit 201, By using a plurality of main storage units 202 and auxiliary storage units 203, it is possible to configure parallel distributed processing by a plurality of computers. Further, as another embodiment, it is also possible to install a plurality of servers for the estimation device 100 so that the plurality of servers share one auxiliary storage unit 203.

The IF unit 204 serves as an interface for transmitting and receiving data to and from other systems and devices, and also provides an interface for receiving various commands and input data (various masters, tables, etc.) from the system operator. The output unit 205 provides a display screen for displaying the processed data, a printing means for printing the data, and the like.

The customer master 206 is a master file that stores customer information. FIG. 3 is a diagram showing an example of a data structure of the customer master 206 according to the embodiment of the present invention. The customer master 206 can include customer ID 301, customer information 302, customer account identification information 303, industry 304, and status 305, but is not limited to these data items and may also include other data items. It is possible.

The customer ID 301 is an identifier that identifies the customer. The customer information 302 indicates customer information such as a customer name, an address, and contact information. The customer account identification information 303 indicates information that identifies one or a plurality of accounts opened by the customer. Depending on the management form, the customer may have opened only one customer account or may have opened multiple accounts at different head offices and branches. When there are a plurality of customer-opened accounts, the estimation device 100 can use the account data after name identification after performing name identification processing on the plurality of accounts.

Industry 304 indicates the industry associated with the customer. The type of industry may be according to the classification set by the Ministry of Economy, Trade and Industry, or may be according to the classification arbitrarily set by an industry group or the like, and is not particularly limited. Status 305 indicates the business condition of the customer (eg, growth, normal, default, etc.) as determined by the financial institution according to predetermined criteria.

Returning to FIG. 2, the deposit / withdrawal detail 207 stores the deposit / withdrawal detail data for the customer account. FIG. 4 is a diagram showing an example of the data structure of the deposit / withdrawal specification 207 according to the embodiment of the present invention. The deposit / withdrawal statement 207 may include an account identifier 401, a transaction date 402, a deposit / withdrawal category 403, an amount 404, a balance 405, and a description 406, but is not limited to these data items. Data items can also be included.

The account identifier 401 is an identifier that identifies a customer account, and is, for example, an identifier based on a bank code, a branch code, a deposit type, an account number, or the like. The transaction date 402 indicates the date on which the deposit transaction or the withdrawal transaction was performed. The deposit / withdrawal category 403 indicates whether the transaction performed is a deposit transaction or a withdrawal transaction. The amount 404 indicates the transaction amount. The balance 405 indicates the balance information of the customer account after the deposit transaction or the withdrawal transaction. Description 406 provides an explanation for a deposit or withdrawal transaction (eg, transfer, transfer, etc.).

Returning to FIG. 2, the learning numerical data 208 stores the frequency distribution data generated by the estimation device 100. The frequency distribution data is based on predetermined first and second period data for each of the variables contained in the deposit and withdrawal statement data of various lender accounts associated with multiple business conditions patterns. Will be generated. The lengths of the first period and the second period may be the same or different. For example, for a company that has already defaulted, the period of poor performance (second period) may be quite short.

FIG. 5 is a diagram showing an example of the data structure of the learning numerical data 208 according to the embodiment of the present invention. The learning numerical data 208 can include industry 501, business situation pattern 502, variable 503, period 504, class 505, and frequency 506, but is not limited to these data items and includes other data items. It is possible.

Industry 501 indicates the type of business to which the loan is made. The classification indicated by industry 501 is the same as the classification indicated by industry 304. The business condition pattern 502 indicates the business condition of the lender (for example, growth, normal, default). Variable 503 indicates a variable included in the deposit / withdrawal detail data. Variables may be, for example, balance, deposit amount, withdrawal amount, number of transactions, number of transfers and amount, number of transfers and amount. The period 504 indicates which period the frequency distribution data is. For example, the period 504 may indicate a first period and a second period, and the first period and the second period may each indicate a specific period (From To). Class 505 indicates a data division when generating frequency distribution data. The frequency 506 indicates the number of values belonging to each class.

Returning to FIG. 2, the training image data 209 stores image data including a histogram (frequency distribution map) generated by using the frequency distribution data stored in the training numerical data 208. FIG. 6 is a diagram showing an example of the data structure of the learning image data 209 according to the embodiment of the present invention. The learning image data 209 can include an industry 501, a business situation pattern 502, a variable 503, an image data 601 and a difference information 602, but is not limited to these data items and may include other data items. It is possible.

Since the industry 501, the business situation pattern 502, and the variable 503 are as described with reference to FIG. 5, detailed description thereof will be omitted. The image data 601 shows image data including histograms generated based on the data of class 505 and frequency 506 of each period indicated by the industry 501, the business situation pattern 502, and the period 504 associated with the variable 503. For example, for a default lender, if period 504 indicates a first period (normal period) and a second period (deterioration period), the image data will include the histogram associated with the first period and the first. Includes a histogram associated with the two periods. The difference information 602 shows the information of the difference value calculated for each class, which is the result of comparing the first histogram and the second histogram.

(Processing flow: image data generation processing and machine learning processing)
FIG. 7 is a diagram illustrating an image data generation process and a machine learning process executed by the estimation device 100. With reference to this processing flow, the estimation device 100 uses a predetermined first period and a first period for each of the variables included in the deposit / withdrawal detail data of the accounts of various lenders associated with a plurality of business situation patterns. Generate each frequency distribution data associated with the data in the second period, and use each generated frequency distribution data to create a first histogram associated with the first period and a second associated with the second period. The process of generating image data including the histogram of 2 will be described. In addition, the process in which the estimation device 100 machine-learns the generated image data will be described with reference to this processing flow.

In S701, the estimation device 100 receives information on the industry and status designated as the processing target. The industry and status may be specified from an operator terminal (not shown) connected to the estimation device 100. The estimation device 100 accesses the customer master 206, reads out the information of the customer account identification information 303 associated with the received industry and status information, and identifies the customer account to be processed.

In S702, the estimation device 100 accesses the deposit / withdrawal detail 207 and reads out the deposit / withdrawal detail data associated with the customer account identified as the processing target.

In S703, the estimation device 100 is based on the predetermined first period and second period deposit / withdrawal detail data for each variable (for example, balance) included in the read deposit / withdrawal detail data. Generate frequency distribution data for. The first period and the second period may be predetermined by the financial institution, and the first period and the second period may be the same length period or different length periods. Good. For example, for a company that has defaulted (default), the first period is one year, which is the normal period before the business situation deteriorates, and the deterioration period until the business situation begins to deteriorate and becomes the default. (N months: N is a natural number) may be the second period. The estimation device 100 stores the generated frequency distribution data in the learning numerical data 208. The value of the class 505 of the frequency distribution data may be arbitrarily determined.

In S704, the estimation device 100 reads out each frequency distribution data stored in the learning numerical data 208, and generates image data including a histogram for each of the industry 501, the business situation pattern 502, and the variable 503. The image data includes a first histogram associated with the first period and a second histogram associated with the second period. Further, the estimation device 100 compares the first histogram and the second histogram, and calculates the difference value for each class. The generated image data and the difference value for each class are associated with the industry 501, the business situation pattern 502, and the variable 503, and are stored in the image data 601 and the difference information 602 of the learning image data 209 by the estimation device 100, respectively. ..

Here, the processes of S703 and S704 will be described with reference to FIG. 9 (a) to 9 (c) are diagrams illustrating a process of generating image data including a first histogram and a second histogram. As shown in FIG. 9A, the estimation device 100 is the first from the value of the time series variation of a variable in the first period (for example, normal period) and the second period (for example, deterioration period). Generate frequency distribution data and second frequency distribution data. The estimation device 100 generates a first histogram of the first period and a second histogram of the second period based on the first frequency distribution data and the second frequency distribution data. The estimation device 100 superimposes the first histogram and the second histogram as illustrated in FIG. 9B, and calculates the difference value for each class. As shown in FIG. 9C, the estimation device 100 generates image data including a first histogram and a second histogram. The generated image data and the calculated difference value are stored in the training image data 209 as described above.

In S705, the estimation device 100 can make artificial intelligence (AI) machine-learn the image data and the difference information stored in the training image data 209. The estimation device 100 can utilize various machine learning methods. For example, the estimation device 100 can use deep learning (deep learning), which is a method of machine learning, and can also use a convolutional neural network (CNN) in the deep learning model.

(Processing flow: Prediction processing)
FIG. 8 is a diagram illustrating a prediction process executed by the estimation device 100. With reference to this processing flow, the estimation device 100 applies the data of the predetermined first period and the second period for each of the variables included in the deposit / withdrawal detail data of the customer account of the customer to be analyzed. By generating image data including a plurality of histograms using each frequency distribution data generated based on the above, and comparing the generated image data with the image data stored in the training image data 209. The process of determining which business situation pattern the customer to be analyzed is likely to correspond to will be described. Further, in this processing flow, AI may perform comparison processing between image data and determination processing as to which business situation pattern the customer to be analyzed is likely to correspond to.

In S801, the estimation device 100 receives the customer ID 301 of the customer designated as the analysis target. The customer ID 301 may be designated from an operator terminal (not shown) connected to the estimation device 100. The estimation device 100 accesses the customer master 206 and identifies the customer account to be analyzed based on the customer account identification information 303 associated with the customer ID 301 of the customer designated as the analysis target. The estimation device 100 also acquires information on the industry 304 associated with the customer ID 301.

In S802, the estimation device 100 accesses the deposit / withdrawal detail 207 and reads out the deposit / withdrawal detail data associated with the customer account identified as the analysis target.

In S803, the estimation device 100 determines the frequency based on the predetermined first period and second period deposit / withdrawal detail data for each variable (for example, balance) included in the read deposit / withdrawal detail data. Generate distribution data. The first period and the second period are the same as the first period and the second period described with reference to FIG. 7. The generated frequency distribution data includes a first frequency distribution data associated with a first period and a second frequency distribution data associated with a second period.

In S804, the estimation device 100 associates with the first histogram and the second period associated with the first period based on the first frequency distribution data and the second frequency distribution data generated in S803. The image data including the second histogram to be generated is generated. The generated image data may be image data as illustrated in FIG. 9 (c). Further, the estimation device 100 compares the first histogram and the second histogram included in the generated image data, and calculates the difference value for each class.

In S805, the estimation device 100 identifies the magnitude of the difference between the first histogram and the second histogram included in the image data associated with the customer to be analyzed generated in S804. This identification process may be performed on an image basis, or may be performed based on the difference value of each class calculated for the image data. The estimation device 100 determines whether or not the magnitude of the identified difference is equal to or greater than a predetermined threshold value. The threshold value may vary depending on the customer's industry and the like. If the magnitude of the difference is greater than or equal to a predetermined threshold (that is, if there is a noticeable change in the business situation between the first period and the second period), the process proceeds to S806 and is less than the threshold. If (that is, there is no noticeable change in business conditions between the first period and the second period), this processing flow ends.

In S806, the estimation device 100 associates the image data generated in S804 with the customer to be analyzed with the corresponding industry 501 and variable 503 (eg, balance) stored in the training image data 209. Compare with one or more image data. The training image data 209 has image data of a plurality of business situation patterns 502 (eg, growth, normal, default) associated with the corresponding industry 501 and variable 503. By this comparison processing, the estimation device 100 uses which image data stored in the training image data 209 to display the difference pattern between the first histogram and the second histogram indicated by the image data associated with the customer to be analyzed. It is possible to determine whether or not the difference pattern between the first and second histograms shown has a high degree of similarity, and predict the business condition of the customer to be analyzed. That is, depending on the business condition pattern associated with the image data determined to have high similarity, the estimation device 100 is in the stage where the customer to be analyzed is in the growth stage, the normal stage, or is moving toward the default. It is possible to predict whether it is in. The criteria for determining whether or not the degree of similarity is high can be arbitrarily set by the lender company, and the criteria can be set for each type of loan recipient.

In the embodiment of the present invention, the estimation device 100 has the difference value for each class calculated for the image data associated with the customer to be analyzed generated in S804 and the correspondence stored in the training image data 209. The difference value for each class included in the difference information 602 of one or more image data associated with the industry 501 and the variable 503 (for example, the balance) is compared. By this comparison processing, the estimation device 100 uses which image data stored in the training image data 209 to display the difference pattern between the first histogram and the second histogram indicated by the image data associated with the customer to be analyzed. It is possible to determine whether or not the difference pattern between the first and second histograms shown has a high degree of similarity, and predict the business condition of the customer to be analyzed.

In the embodiment of the present invention, the estimation device 100 provides the artificial intelligence (AI) with the image data generated in S804 and associated with the customer to be analyzed, and compares it with the image data machine-learned in advance. The pattern of the difference between the first and second histograms, which is shown by the image data associated with the customer to be analyzed, is the first histogram and which of the machine-learned image data is shown. By determining whether the difference pattern of the second histogram has a high degree of similarity, it is possible to predict the business situation of the customer to be analyzed. That is, the estimation device 100 is likely to correspond to the business condition pattern (for example, growth, normal, default) associated with the image data determined by AI to have a high degree of similarity to the business condition of the customer to be analyzed. Can be determined. Further, the estimation device 100 can further provide the AI with the difference value for each class calculated in S804 when the AI is made to perform the comparison. In such a case, the AI is the image data and the difference value. The similarity can be determined based on both.

In the embodiment of the present invention, the estimation device 100 uses the first frequency distribution data associated with the first period and the second frequency distribution associated with the second period among the frequency distribution data generated in S803. By comparing with the data, it is possible to infer whether the customer to be analyzed is in the growth stage or the decline stage. More specifically, the estimation device 100 divides the second average value calculated based on the second frequency distribution data by the first average value calculated based on the first frequency distribution data. If a value greater than 1 is obtained, it can be determined to be in the growth stage, and if a value smaller than 1 is obtained, it can be determined to be in the decline stage. It should be noted that the average value may be the actual data of the variables (for example, balance, deposit amount, withdrawal amount, etc.) of the first period and the second period.

By the processing as described above, the estimation device 100 analyzes the image data based on the dynamic data of the account of the specific lender based on the accumulated image data or the machine-learned image data, thereby causing the financial numerical data. It will be possible to predict large fluctuations in business conditions, especially the occurrence of defaults, at an earlier stage than before using.

Although the principles of the present invention have been described above with reference to exemplary embodiments, those skilled in the art will appreciate that various embodiments that change in configuration and details can be realized without departing from the gist of the present invention. You will understand. That is, the present invention can be implemented as, for example, a system, an apparatus, a method, a program, a storage medium, or the like.

100 Estimator 110 Banking system 120 Customer terminal 130 ATM
140 Other bank system 201 Control unit 202 Main storage unit 203 Auxiliary storage unit 204 Interface (IF) unit 205 Output unit 206 Customer master 207 Deposit / withdrawal details 208 Learning numerical data 209 Learning image data 210 Bus

Claims (11)

It ’s an estimation device,
Read the deposit / withdrawal detail data associated with the customer specified for analysis,
For the variables included in the read deposit / withdrawal detail data, the first frequency distribution data associated with the first period and the second frequency distribution data associated with the second period are generated.
A first image containing a first histogram associated with the first period and a second histogram associated with the second period, based on the first frequency distribution data and the second frequency distribution data. Generate data and
The magnitude of the difference between the first histogram and the second histogram included in the first image data is identified, and it is determined whether or not the magnitude of the identified difference is equal to or greater than a predetermined threshold value. And
The first image data and the second image data are compared on the condition that the magnitude of the difference is determined to be equal to or larger than a predetermined threshold value, and the first image data is included in the first image data. Based on a predetermined criterion, which image data of the second image data has a high degree of similarity to the difference pattern between the first histogram and the second histogram. According to the business condition pattern associated with the second image data determined to have a high degree of similarity, the customer designated as the analysis target is in the growth stage, the normal stage, or the default. An estimation device configured to predict the business situation of the customer by determining whether it is in the oncoming stage . The estimation device according to claim 1, wherein the predetermined standard is determined for each type of business of the lender. The estimation device according to claim 1 , wherein the comparison and the determination are performed by artificial intelligence (AI). The estimation according to claim 1, further configured to compare the first histogram and the second histogram included in the first image data and calculate a first difference value for each class. apparatus. The first difference value for each class calculated for the first image data is compared with the corresponding second difference value associated with the second image data, and the first image data is obtained. By determining which image data of the second image data has a high degree of similarity to the pattern of the difference between the first histogram and the second histogram included. The estimation device according to claim 4, further configured to predict the business situation of the customer. The estimation device according to claim 5, wherein comparing the first difference value with the second difference value and determining whether the similarity is high is performed by artificial intelligence (AI). The estimation device according to claim 1, wherein discriminating the magnitude of the difference between the first histogram and the second histogram is performed based on the first image data. Discriminating the magnitude of the difference between the first histogram and the second histogram is further configured to be performed based on the difference value of each class calculated for the first image data. The estimation device according to claim 1. Claim 1 is further configured to infer whether the customer is in a growth stage or a decline stage by comparing the first frequency distribution data with the second frequency distribution data. The estimation device described. An estimation method performed by an estimation device,
Reading the deposit / withdrawal detail data associated with the customer specified for analysis, and
To generate the first frequency distribution data associated with the first period and the second frequency distribution data associated with the second period for the variables included in the read deposit / withdrawal detail data.
A first image containing a first histogram associated with the first period and a second histogram associated with the second period, based on the first frequency distribution data and the second frequency distribution data. To generate data and
The magnitude of the difference between the first histogram and the second histogram included in the first image data is identified, and it is determined whether or not the magnitude of the identified difference is equal to or greater than a predetermined threshold value. To do and
The first image data and the second image data are compared on the condition that the magnitude of the difference is determined to be equal to or larger than a predetermined threshold value, and the first image data is included in the first image data. Based on a predetermined criterion, which image data of the second image data has a high degree of similarity to the difference pattern between the first histogram and the second histogram. According to the business condition pattern associated with the second image data determined to have a high degree of similarity, the customer designated as the analysis target is in the growth stage, the normal stage, or the default. An estimation method that includes predicting the business situation of the customer by determining whether or not the customer is in the on-going stage . A program that causes an estimation device to execute the method according to claim 10, when executed by a processor.

Images