JP7171482B2 - Business Exceptional Case Extraction Support System and Business Exceptional Case Extraction Support Method - Google Patents

Description

TECHNICAL FIELD The present invention relates to technology for assisting the extraction of out-of-business cases.

In Patent Document 1, in order to improve the efficiency of verification of materials in the examination department, a combination of verification items associated with materials is stored, and necessary verification items for new materials are narrowed down and presented. Techniques to assist verification are described.

Japanese Patent Application Laid-Open No. 2009-238173

When verifying the accuracy of business results for applications, that is, when detecting business error cases through verification, it is difficult to verify all applications from the perspective of human resources. There is no choice but to verify the department's application form. The problem is that the detection rate of business error cases is low when some applications are randomly selected from all applications for verification.

Patent Document 1 discloses a method for streamlining verification by narrowing down verification items. Not applicable.

The present invention provides a business exception case extraction support system capable of increasing the detection rate of business error cases when extracting and verifying some application forms from all application forms, and a business exception case extraction support system. An object of the present invention is to provide an extraction support method.

A business failure case extraction support system according to the present invention is a business failure case extraction support system for extracting failure cases of examination business of an application, comprising nominal scale data included in the application, and the nominal scale data. generating first 1/0 data for the nominal scale data based on the category values corresponding to the ordinal/interval/ratio scale data contained in the application and generating a second 1/0 data for the ordinal/interval/ratio scale data based on the corresponding category values and the first 1/0 data and the second 1/0 data for the application form; a 1/0 data creation unit that generates a 1/0 table containing data, and a cluster analysis that performs cluster analysis based on the distance between the application forms in the generated 1/0 table and calculates the clusters of the application forms. , and is configured as a business exception case extraction support system.

According to the present invention, it is possible to increase the detection rate of business error cases when extracting and verifying some application forms from all the application forms.

It is an example of a system configuration diagram. It is an example of a business support server. This is an example of a business server. It is an example of the overall processing flow of the out-of-business case extraction system. It is an example of the overall processing flow of the out-of-business case extraction system. It is an example of the overall processing flow of the out-of-business case extraction system. It is an example of the processing flow of a business support server. It is an example of the processing flow of a business support server. It is an example of application form information data. It is an example of category definition for nominal scale data. Examples of category definitions for ordinal, interval, and ratio scale data. It is an example of 1/0 data based on category definition. This is an example of an application-cluster correspondence table. It is an example of cluster information. It is an example of distance information between major clusters. It is an example of a cluster information confirmation screen. It is an example of an out-of-work case extraction result screen. It is an example of the change difference extraction result screen of last year and this year.

Embodiments of the present invention will be described below with reference to the drawings. The following description and drawings are examples for explaining the present invention, and are appropriately omitted and simplified for clarity of explanation. The present invention can also be implemented in various other forms. Unless otherwise specified, each component may be singular or plural.

The position, size, shape, range, etc. of each component shown in the drawings may not represent the actual position, size, shape, range, etc., in order to facilitate understanding of the invention. As such, the present invention is not necessarily limited to the locations, sizes, shapes, extents, etc., disclosed in the drawings.

In the following description, various types of information may be described using expressions such as “table” and “list”, but various types of information may be expressed in data structures other than these. "XX table", "XX list", etc. are sometimes referred to as "XX information" to indicate that they do not depend on the data structure. When describing identification information, expressions such as “identification information”, “identifier”, “name”, “ID”, and “number” are used, but these can be replaced with each other.

When there are a plurality of components having the same or similar functions, they may be described with the same reference numerals and different suffixes. However, if there is no need to distinguish between these multiple constituent elements, the subscripts may be omitted in the description.

Also, in the following description, there are cases where processing performed by executing a program is described, but the program is executed by a processor (for example, CPU, GPU) to appropriately perform the specified processing using storage resources ( For example, a memory) and/or an interface device (for example, a communication port) or the like is used, so processing may be performed by a processor. Similarly, a main body of processing executed by executing a program may be a controller having a processor, a device, a system, a computer, or a node. The subject of the processing performed by executing the program may be an arithmetic unit, and may include a dedicated circuit (for example, FPGA or ASIC) that performs specific processing.

A program may be installed on a device, such as a computer, from a program source. The program source may be, for example, a program distribution server or a computer-readable storage medium. When the program source is a program distribution server, the program distribution server may include a processor and storage resources for storing the distribution target program, and the processor of the program distribution server may distribute the distribution target program to other computers. Also, in the following description, two or more programs may be implemented as one program, and one program may be implemented as two or more programs.

FIG. 1 shows a configuration of a business system 1000 in which a business support server 10, a business server 20, and a business client terminal 30 are interconnected via a network 5, as an example of a business system. A business operator 1 uses a business client terminal 30 to utilize a business system 1000 .

FIG. 2 is a configuration example of the business support server 10. As shown in FIG. The business support server 10 is composed of a storage unit 101 and a calculation unit 102 . The storage unit 101 is configured by, for example, a storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive) as hardware. Further, the computing unit 102 is configured by, for example, a computing device having a CPU (Central Processing Unit), RAM (Random Access Memory), and ROM (Read Only Memory) as hardware.

The storage unit 101 stores application form information data 11, category definitions 12 of nominal scale data, category definitions 13 of order, interval, and ratio scale data, 1/0 data 14 based on the category definitions, and application form-cluster correspondence. It consists of Table 15, cluster information 16, and distance information 17 between major clusters.

The calculation unit 102 includes a 1/0 data creation unit 1A based on the category definition, a cluster analysis unit 1B, a business exception case extraction unit (business exception case extraction unit) 1C, and a business cluster corresponding to the application form 2. It is composed of an out-of-case extraction unit (cluster out-of-business case extraction unit) 1D and an inter-cluster information difference extraction unit 1E.

Each unit included in the calculation unit 102 is implemented by executing a program. For example, the function of each part of the arithmetic unit 102 is realized by the CPU reading out and executing a program from the ROM. The program may be provided to the business support server 10 by being read from a storage medium such as a USB (Universal Serial Bus) memory or downloaded from another computer via a network.

FIG. 3 is a configuration example of the business server 20. As shown in FIG. The business server 20 comprises a storage unit 201 for storing application information data 21 having the same content as the application information data 11 of the business support server 10 and a calculation unit 202 comprising a business result recording unit 22 . As hardware, the storage unit 201 and the calculation unit 202 are configured by a conventionally known general computer, like the business support server 10 .

4, 5 and 6 show three overall sequence examples.

FIG. 4 is an example of an overall sequence for analyzing and processing application form information data and displaying cluster information and results of extracting non-work cases. Each step of the sequence will be described below.

The business operator 1 calls the business result recording unit 22 of the business server 20 via the business client terminal 30 , and the business result recording unit 22 records the application form information data 21 . The application form information data 21 has the same format as the application form information data 11 in FIG. For example, the business operator 1 inputs the contents of one application form to 1102 to 1110, . If they match, enter “OK” in the examination result 1111 . This operation is repeated by the number of applications, and the work result recording unit 22 creates the same number of records as the number of applications. (S201)
The business support server 10 receives the application information data 21 of the business server 20, copies it to the application information data 11 (S100), and analyzes and processes the application information data 11 by the processes of FIGS. (S101-S107)
The business support server 10 displays a cluster information confirmation screen 300 to the business operator 1 via the business client terminal 30 . The business operator 1 checks the displayed cluster information confirmation screen 300 to check what patterns (clusters) are included in the application form and what characteristics each cluster has. can be done. Also, by checking the displayed tree diagram (dendrogram) 301, it is possible to check the similarity between clusters. These allow the business operator 1 to grasp the overall trend of the application forms. (S110)
The business support server 10 displays a business exception case extraction result screen 310 to the business operator 1 via the business client terminal 30 . The business operator 1 can confirm rare business cases 311, business cases 312 with many errors, and business cases 313 with few major errors that are displayed on the business exception case extraction result screen 310. FIG. (S111) A rare business case 311, a business case 312 with many errors, and a business case 313 with few major errors are determined in the processing of FIGS.

Two utilization examples of the business exception case extraction result screen 310 displayed in S111 will be described below.
The first application example will be explained. Operator 1, who is in charge of checking whether there are any business errors, is assumed to have many business errors in rare application forms with rare application content. By preferentially checking from the viewpoint of the cluster feature 164, it is possible to efficiently check whether there is an operational error.

A second application example will be described. Examination result 1111 for the application form = NG judgment (no judgment) Business operator 1 in charge of checking whether there is any omission, the examination result = NG application form and similar examination result = Since there is a high possibility that it should be NG, the application form information data 11 corresponding to the business case 312 with many errors on the screen 310 is preferentially checked mainly from the viewpoint of the cluster characteristics 164. Then, it is possible to check whether or not there is anything that should be judged as the examination result 1111 = NG.

FIG. 5 is an example of an overall sequence in which the application form information data 11 is analyzed and processed, and the relevant part of the result of the out-of-work case extraction is displayed for the new application form 2 . Each step of the sequence will be described below.

First, the processes of S201, S100, and S101 to S107 similar to those in FIG. 4 are performed.

The business operator 1 inputs the information of the new application form 2 (the information of 1102 to 1110, . . . of the application form information data 11) to the business support server 10 via the business client terminal 30 (S112).

The business support server 10 extracts the information input in S112 in three cases (a rare business case 311, a business case 312 with many errors, and a business case 313 with few major errors). , and only the applicable cases (clusters) are displayed on the business exception case extraction result screen 310 of the business client terminal 30 (in the example of FIG. 17, only the applicable cases (clusters) Show). (S113)
Two utilization examples of the information displayed on the out-of-work case extraction result screen 310 will be described below.
The first application example will be explained. It is assumed that there are experienced staff members who are familiar with the examination rules and novice staff members who are not familiar with the examination rules among the staff members who examine applications. The operator 1 who distributes the application form to the experienced staff and the novice staff inputs the application form 2 information in S112. In S113, when the business support server 10 determines and displays a rare business case 311 or a business case 312 with many errors, the application form 2 is sorted to the veteran staff, and the feature 164 of the corresponding cluster is displayed. Provide veteran staff. In S113, when the business support server 10 determines that it corresponds to the common business case 313 with few major errors and displays it, it distributes the application form 2 to the novice staff. By optimizing the allocation of applications in this way, the entire examination process can be made more accurate and faster.

A second application example will be described. Suppose operator 1 wishes to detect a unique application form among all application forms. In this case, in S113, when the business support server 10 determines that the business failure case extraction result screen 310 corresponds to the rare business case 311, it can be determined that a peculiar application form has been detected.

FIG. 6 is an example of an overall sequence in which the group of application forms for the previous year and the group of application forms for the current year are analyzed and processed, respectively, the cluster information 16 is created, and the difference between them is extracted and displayed. Each step of the sequence will be described below.

First, the same processing of S201, S100, and S101 to S107 as in FIG. 4 is performed for the group of application forms of the previous year and the group of application forms of the current year.

Next, the business support server 10 extracts the cluster information 16 for the previous year and the cluster information 16 for the current year obtained from the application records belonging to the clusters (1102 to 1110 of the application information data 11, . . . ). ) are the same or very similar, the clusters for the previous year and the clusters for the current year are associated with each other (such as "Cluster 1 last year" and "Cluster 1 this year" in the cluster change difference 321 in FIG. 18). ). An example in which the differences are displayed in association is the change difference extraction result screen 320 of FIG. 18 between the previous year and the current year. (S114) The similarity of the records is very high, for example, when a predetermined criterion is satisfied (for example, when 80% or more of the total number of items match).

Three utilization examples of the information displayed on the change difference extraction result screen 320 will be described below.

The first application example will be explained. Assume that there is an operator 1 who wants to check the changes in the application form and examination results when there is no significant change in the application form content or the examination work between the previous year and the current year. Operator 1 uses the cluster 1 change difference 321, the cluster 4 change difference 322, and the current year new cluster 323 on the change difference extraction result screen 320 of the previous year and the current year displayed in S114 to extract the difference between the previous year and the current year. It is possible to explore the reasons for the change in

A second application example will be described. Suppose that there is an operator 1 who wants to grasp changes in application contents and examination work before and after revision of the system. By performing the same processing as in FIG. 6 on the application forms before and after the revision of the system, the operator can check the change difference extraction result screen 320 displayed in S114 to obtain the group of application forms before and after the revision of the system. It is possible to understand changes in application content and examination work after revision.

A third application example will be described. Suppose that there is an operator 1 who wants to grasp monthly changes in application forms and examination work. By performing the same processing as in FIG. 6 for the group of monthly application forms, the operator 1 can check the change difference extraction result screen 320 displayed in S114, thereby confirming the content of the monthly application forms and examination work. change can be grasped.

7 and 8, the business support server 10 processes the application form information data 11 to create 1/0 data 14 based on the category definition (S101 to S104), performs cluster analysis (S105), cluster information 164 is created (S106-S107) processing flow. Each step of the flow will be described below.

The business support server 10 receives a copy of the application information data 21 from the business server 20 and stores it in the application information data 11 (S101).

Here, the types of each field (1101 to 1110, . Each field of data is of two types.

The first is data that has the meaning of whether a certain value is different from or the same as another, and is called "nominal scale data". For example, fields 1101, 1103, 1106, 1107, 1108, and 1110 correspond to this. In general, categories such as gender, cohabitation/separation, etc. are examples of nominal scale data.

The second type is data having the meaning of order, interval, and ratio of values, and is called "order, interval, ratio scale data". For example, fields 1102, 1104, 1105 and 1109 correspond to this. In general, years and months are examples of ordinal and interval scale data, money is ratio scale data, temperature is ratio scale data, and qualification grades are ordinal scale data.

The business operator 1 accesses the business support server 10 via the business client terminal 30, and the 1/0 data creation unit 1A creates the category definition 12 of the nominal scale data.

That is, according to the operation of the business operator 1, the 1/0 data creation unit 1A transcribes the category value 122 of the category definition 12 of the nominal scale data from the design document of the application form information data 21 of the business server 20. For example, the category value 122 column corresponding to the applicant category 1103, which is the field name of the application form information, means that the applicant category 1103 can take four values "applicant categories A to D."

Furthermore, according to the operation from the business operator 1, the 1/0 data creation section 1A, if there is a superordinate category value that summarizes one or more category values 122 for the category value 122 of each field name, converts it to a superordinate category value. Described in the category value 123. For example, if there are four category values of applicant categories "eldest son", "eldest daughter", "second son", and "second daughter", which are field names of application information, these can be associated with the upper category 123 "child". can.

By the operation from the business operator 1, the 1/0 data creation unit 1A performs these processes on all nominal scale data fields (1101, 1103, 1106, 1107, 1108, 1110, . . . ) of the application form information data 21. to create a category definition 12 for nominal scale data. (S102)
The business operator 1 accesses the business support server 10 via the business client terminal 30, and the 1/0 data creation unit 1A creates category definitions 13 for order, interval, and ratio scale data.

By an operation from the business operator 1, the 1/0 data creation unit 1A determines the correspondence between the value 132, which is the delimiter of the business, and the category value 133 corresponding to the value 132, from a predetermined business viewpoint determined by the business operator 1. Decide from and enter. For example, the applicant date of birth field 1102 indicates that it is divided into 4 categories (category value 133) corresponding to the value 132 of the year of birth. This allows order, interval, and ratio scale data to be associated with category values 133 .

By the operation from the business operator 1, the 1/0 data creation unit 1A performs these processes on all the order, interval, ratio scale data fields (1102, 1104, 1105, 1109, . . . ) of the application form information data 21. A category definition 12 for nominal scale data is created repeatedly. (S103)
The 1/0 data creation unit 1A based on the category definition of the business support server 10 refers to the category definition 12 of the nominal scale data and the category definition 13 of the order, interval, and ratio scale data, and categorizes the application form information data 11 into categories. It is converted to a value (1 or 0) and stored in 1/0 data 14 based on the category definition.

A specific example in the case of nominal scale data is shown. The 1/0 data creation unit 1A of the business support server 10 acquires the value “applicant category A” of the record 1 (112) of the applicant category 1103 of the application form information data 11, Refer to line 1201 corresponding to "Applicant Category A" to acquire the upper category value "Applicant Upper Category 1" of "Applicant Category A". Next, the 1/0 data creation unit 1A creates a record 1 corresponding to "applicant classification A" for the record 1 column (143) of the applicant classification 1103 row of the 1/0 data 14 based on the category definition. "1" is stored in the value of column (143), and "0" is stored in the value of record 1 column (143) corresponding to the other "applicant categories B to D". Furthermore, the 1/0 data creation unit 1A sets the value of the record 1 column (143) corresponding to the "applicant upper category 1" to "1" for the record 1 column (143) of the applicant upper category 1121 row. is stored, and "0" is stored in the value of the other record 1 column (143) corresponding to "applicant upper division 2" (first 1/0 data).

Next, specific examples for order, interval, and ratio scale data will be shown. The 1/0 data creation unit 1A of the business support server 10 acquires the value "1950/01/01" of the record 1 (112) of the applicant date of birth 1102 of the application form information data 11, Refer to line 1303 corresponding to "1950/01/01" in the scale data category definition 13 to obtain the category value "applicant's date of birth division 3". Next, in the 1/0 data 14 based on the category definition, for the record 1 column (143) of the applicant's date of birth 1102 row, the record 1 column (143) corresponding to "applicant date of birth classification 3" "1" is stored in the value of , and "0" is stored in the value of record 1 column (143) corresponding to other "applicant's date of birth classification 1, 2, 4" (second 1/ 0 data).

The 1/0 data creation unit 1A of the business support server 10 performs these operations on all fields 111 other than the "examination result field" for all records of the application form information data 11, and creates 1 data based on the category definition. /0 data 14 is created. (S104)
The cluster analysis unit 1B of the business support server 10 divides the records (143, 144, . . . ) of the 1/0 data 14 based on the category definition into A round-robin distance is calculated, and cluster analysis is performed based on the calculated distance. Note that the distance is calculated using the L1 distance. For cluster analysis, hierarchical cluster analysis or non-hierarchical cluster analysis is performed using a machine learning software library or the like.

The cluster analysis unit 1B of the business support server 10 stores the results of the cluster analysis in the application form-cluster correspondence table 15. FIG. FIG. 13 is an example of the application form-cluster correspondence table 15 when the cluster analysis result indicates that record 1 with application form ID 00001 belongs to cluster 1 and record 2 with application form ID 00002 belongs to cluster 2. . (S105)
The cluster analysis unit 1B of the business support server 10 refers to the examination result 1111 of the application form information data 11 and the application form-cluster correspondence table 15 to determine the number of belonging records 161 of the cluster information 16 and the main cluster determination result. 162 and the number 163 of examination results=NG are calculated and stored.

Each calculation method will be described below.

The cluster analysis unit 1B counts and stores the number of records that match the belonging cluster 1501 row of the application form-cluster correspondence table 15 as the belonging record number 161. FIG. For example, the cluster analysis unit 1B counts the number of records with "Cluster 1" in the row 1501 of the belonging cluster for cluster 1 (in the example of FIG. 13, one is counted for record 1) and stored. A value of “2000” in the number of records belonging to cluster 1 161 of the cluster information 16 indicates that the application form-cluster correspondence table 15 has 2000 records of “cluster 1” in the belonging cluster 1501 row.

As the main cluster determination result 162, the cluster analysis unit 1B stores a value "○" if the number of belonging records 161 is equal to or greater than a predetermined threshold value (eg, number of belonging records = 10), and stores "X" if less than. to store As a result, clusters having the number of belonging records 161 equal to or greater than the threshold can be identified (○ or X). In other words, it is possible to extract (○) a common, major, similar application form group (cluster).

The cluster analysis unit 1B extracts the corresponding application form ID 1101 from the application form-cluster correspondence table 15 for each cluster as examination result = NG number 163, and examines the application form information data 11 corresponding to the extracted application form ID 1101. The number of NG in the result row 1111 is counted and stored. For example, the cluster analysis unit 1B extracts the value "00002" of the application form ID 1101 for which the value of the belonging cluster 1501 of the application form-cluster correspondence table 15 is "cluster 2" for cluster 2, The value "NG" of the examination result 1111 in the 113 column whose value of the application form ID 1101 is "00002" is referenced, and one "NG" is counted. The cluster analysis unit 1B stores the counted value in the number 163 of examination result=NG. By performing these operations for each cluster in the cluster information 16, the number of examination results=NG for each cluster can be obtained. That is, it can be understood whether the application form group (cluster) is a group with few business errors or a group with many business errors. (S106)
The business operator 1 accesses the business support server 10 via the business client terminal 30, and the cluster analysis unit 1B refers to the application form-cluster correspondence table 15 and the application form information data 11 to obtain the cluster information 16. are examined and input the cluster features 164 of .

An example of searching and inputting two cluster features 164 will now be described.

A first example will be described. The business operator 1 focuses on the cluster 1 of the cluster information 16, refers to the application-cluster correspondence table 15 to identify the corresponding record 1, refers to the value of the record 1 (112) of the application information data 11, While utilizing business knowledge, the document of "regular child support application pattern" in the cluster feature 164 of the cluster 1 of the cluster information 16 is composed, and the cluster analysis unit 1B inputs it. If there are a plurality of corresponding records, the cluster analysis unit 1B uses a statistical tool or the like to refer to the statistical information of the values of the plurality of records, and composes and inputs the cluster feature 164 . Further, the business operator 1 refers to the number of belonging records "2000" in the cluster information 16, and since the number of belonging records is larger than that of other clusters, the "typical" document in the cluster feature 164 of the cluster 1 and the cluster analysis unit 1B additionally inputs.

A second example will be described. The business operator 1 focuses on the cluster 2 of the cluster information 16, refers to the application-cluster correspondence table 15 to identify the corresponding record 2, refers to the value of the record 2 (113) of the application information data 11, The feature of record 2 is that "applicant's date of birth 1102 and applicant classification 1103 are inconsistent", and the examination result is NG. 164, and input by the cluster analysis unit 1B. As a result, the reason for the large number of examination results=NG in cluster 2 can be described in the cluster feature 164 . (S107)
FIG. 9 is an example of the application form information data 11. As shown in FIG. The application form information data 21 has the same contents. The values of each record 112, 113, . . . corresponding to the field name 111 are stored. The application form information data 11 is, for example, the contents of an application form when a resident submits an application form for certification or the like to a public institution. After the application documents are submitted to the public institution in writing or online, the content is entered by the staff of the public institution. Record 1 (112) and record 2 (113) each correspond to one application document.

For example, in the example of record 1 (112), the application form ID "00001" is the number assigned to the application form, and from the applicant's date of birth 1102, the deduction category 1110, ... is the information described in the application form. is. The application is examined by a staff member of the public institution, and if the application satisfies the examination criteria, the examination result 1111 is entered as "OK", and if the application does not meet the examination criteria, the examination result 1111 is ". NG" is entered. The examination result does not have to be binary such as "OK" or "NG". For example, the benefit amount based on the application may be "10,000 yen", "20,000 yen", or "30,000 yen", or the certified classification "Division A", "Division B", or "Division C". good too.

FIG. 10 is an example of category definition 12 of nominal scale data. Nominal scale data is data that has the meaning of whether the value is different from others or the same. , cohabitation/separate division 1108, and deduction division 1110 correspond to this. For example, the application category field 1103 indicates that four category values 122 "applicant category A" to "applicant category D" can be taken. The upper category value 123 is a category value that bundles one or more category values 122, and is a higher category that bundles the category values 122 "Applicant Category A", "Applicant Category B", and "Applicant Category C". A value of 123 means "applicant superclass 1".

FIG. 11 is an example of category definitions 13 for ordinal, interval, and ratio scale data. It is data having the meaning of the order, intervals, and ratios of values. Years and months are ordinal and interval scale data, monetary amounts are ratio scale data, temperatures are ratio scale data, and qualification grades are ordinal scale data.

In the example of applicant date of birth 1102, it is divided into four value ranges as indicated by value 132, which are associated with category values 133 respectively (1301, 1302, 1303, 1304).

FIG. 12 is an example of 1/0 data 14 based on category definitions. , of the application form data 11 are converted (S104) using the category definition 12 of nominal scale data and the category definition 13 of order, interval, and ratio scale data .

Application form ID 1101 , applicant date of birth 1102 , applicant classification 1103 , applicant upper level classification 1104 , . 1122 corresponds to the upper category value 123 of the category definition 12 of the nominal scale data. The category values 142 list the category values 122 and upper category values 123 of the category definition 12 for nominal scale data and the category values 133 of the category definition 13 for ordinal, interval, and ratio scale data.

For example, the applicant's date of birth 1102 of record 1 (143) indicates that it falls under the “applicant's date of birth classification 3” (value=“1”).

FIG. 13 is an example of an application form-cluster correspondence table 15 obtained as a result of cluster analysis. In cluster analysis, each record is classified into one of clusters using the distance between records. For example, the belonging cluster 1501 of record 1 with application form ID 1101="00001" indicates that it is "cluster 1".

FIG. 14 is an example of cluster information 16 indicating the characteristics of each cluster obtained as a result of cluster analysis.

The number of belonging records 161 indicates the number of records belonging to the cluster. For example, cluster 1 has 2000 records and cluster 2 has 5 records. From this, the out-of-work case extraction unit 1C determines that there are 2000 applications belonging to cluster 1 and 5 applications belonging to cluster 2. FIG. From this, the out-of-work case extraction unit 1C can infer that the application forms belonging to cluster 1 are very common major application forms, and the application forms belonging to cluster 2 are rare case application forms. If a predetermined threshold is satisfied, it can be determined that the application is a very common major application. The above threshold can be arbitrarily determined according to the content of the work.

The main cluster determination result 162 is obtained by assigning “◯” and “X” depending on whether or not the number of belonging records 161 is equal to or greater than a certain threshold. For example, the out-of-work case extraction unit 1C sets the number of records belonging to cluster 1 to “2000” because it is greater than the threshold value of 10, so it sets “○”, and the number of records belonging to cluster 2, “5”, because it is smaller than the threshold value of 10. X”. Focusing only on clusters marked with "○" allows a list of major application patterns, and focusing on clusters marked with "X" allows a list of application patterns for rare cases.

The examination result=NG number 163 is obtained by counting the number of examination results 1111="NG" in the application information data 11 among the records belonging to the cluster. The out-of-work case extraction unit 1C counts how many application forms judged as "NG" in the examination are included in the cluster, and outputs the result. Incidentally, as described in the explanation of FIG. 9 above, the examination result 1111 may be, for example, the benefit amount "10,000 yen", "20,000 yen", or "30,000 yen" based on the application. In such a case, instead of examination result = NG number 163, examination result = number of payment amount "10,000 yen", examination result = number of payment amount "20,000 yen", examination result = payment amount You may create a column such as the number of 30,000 yen and count each.

In the cluster characteristics 164, the number of belonging records 161, the main cluster judgment result 162, the number of examination results = NG 163, and the document written and input by the business operator 1 with reference to the business manual etc. and is stored by the non-business case extraction unit 1C.

FIG. 15 is an example of the distance information 17 between major clusters. The main inter-cluster distance information 17 is part of the information generated intermediately in the cluster analysis processing S105. The inter-cluster information difference extracting unit 1E calculates the distance between each cluster for the clusters with the main cluster determination result=“◯” in the cluster information 16, and stores the result. In the example of FIG. 15, when viewed from cluster 1, cluster 5 is further away than cluster 3 (3<10). That is, it means that the similarity between the group of application forms belonging to cluster 1 and the group of application forms belonging to cluster 3 is higher than the similarity between the group of application forms belonging to cluster 1 and the group of application forms belonging to cluster 5. do.

FIG. 16 is an example of a cluster information confirmation screen 300 displayed on the business client terminal 30 operated by the business operator 1 . A tree diagram (dendrogram) 301 at the top of the screen is a tree structure drawn using the distance information 17 between major clusters, and is created by a software tool or the like. It shows that the distance between the main cluster 1 and the main cluster 3 is short, and the main cluster 5 is a little away from the main cluster 1 and the main cluster 3. This allows the business operator 1 to confirm what kind of relationship each main cluster has in terms of distance.

The cluster information 16 at the bottom of the screen displays the cluster information 16 of the business support server 10 . The application form has patterns such as cluster 1, cluster 2, . . . , and the characteristics of each pattern can be listed. The screen is displayed by the inter-cluster information difference extraction unit 1E (or the cluster analysis unit 1B). In the example of the overall sequence shown in FIG. 5, the cluster analysis unit 1B uses the above-described method to determine the cluster to which the input new application record belongs, and the cluster to which the new application record belongs. The number of records and the number of records for which the screening result was rejected may be displayed on the display unit. As a result, the cluster to which the new application record belongs, the number of records belonging to the cluster, and the number of records for which the examination result was rejected can be confirmed at a glance.

FIG. 17 is an example of a business exception case extraction result screen 310 displayed on the business client terminal 30 operated by the business operator 1 .

The cluster work failure case extraction unit 1D extracts the main cluster determination result 162 = "X" from the cluster information 16 of the work support server 10 as the rare business case 311 at the top of the screen, and extracts the work failure case extraction result. Output to screen 310 . Rare business cases (there may be cases where the business operator accidentally recorded the application form information in S201) are displayed here.

The cluster work failure case extraction unit 1D extracts a business case 312 with many errors in the middle part of the screen, from the cluster information 16 of the business support server 10, the ratio of the examination result = NG number 163 to the number of belonging records 161 is greater than or equal to a predetermined threshold. line is extracted and output to the business exception case extraction result screen 310 . As a result of the extraction, the Cluster Features 164 column displays the features of error-prone business cases.

The cluster-out-of-work case extraction unit 1D extracts a business case 313 with few common major errors at the bottom of the screen from the cluster information 16 of the business support server 10. Rows in which the number 163 of examination results=NG for the number of records 161 is equal to or less than a predetermined threshold are extracted and output to the out-of-business case extraction result screen 310 . For example, the extraction result of S113 is displayed as a case such as "Cluster 5" indicated by shading. Here, it is expected that typical common application patterns with few application errors and business errors, and therefore few examination results (NG) will be displayed.

FIG. 18 is an example of a change difference extraction result screen 320 between the previous year and the current year displayed on the business client terminal 30 operated by the business operator 1 .

The cluster out-of-work case extraction unit 1D (or cluster analysis unit 1B) extracts cluster 1 change difference 321 from the cluster information 16 for the previous year's application form group and the cluster information 16 for the current year's application form group. Clusters with high values are extracted and displayed in parallel (Cluster 1 previous year row and Cluster 1 current year row), and the difference row is displayed. Note that portions where the difference is larger than a predetermined threshold are displayed with thick frames. The cluster 4 change difference 322 is similar. Thus, in FIG. 18, the 1/0 data creation unit 1A creates the first application form group (for example, the previous year's application form group) and the second application form group (for example, the current year's application form group). The cluster analysis unit 1B generates a 1/0 table for each of the first application group and the second application group, and calculates the number of records belonging to each cluster and the number of records for which the examination result was negative, and are counted, and the difference between the number of records belonging to the cluster and the number of records for which the examination result was negative, between the first application form group and the second application form group, is displayed on the display unit. As a result, the business operator 1 can grasp the changes in the overall characteristics of the application forms between the previous year and the current year.

The cluster out-of-work case extraction unit 1D (or the cluster analysis unit 1B) has this year's new cluster 323 in the current year's cluster information 16, but the previous year's cluster information 16 contains the same or very similar cluster Show what is missing. This indicates that an application pattern that did not exist in the previous fiscal year has newly emerged this fiscal year.

In the example of FIG. 18, the change difference between the previous year and the current year is extracted. It is also possible to extract differences, extract differences for each region, and extract differences before and after the revision of the legal system. This makes it possible to grasp changes in application patterns by month or year, changes in application patterns due to social trends, regional characteristics of application patterns, changes in application patterns due to revisions of legal systems, and the like.

As described above, in this embodiment, for application form information in which nominal scale data and ordinal/interval/ratio scale data are mixed, a table is created in advance that associates the values of the order/interval/ratio scale data with the category values. Keep By converting the nominal scale data in the application form information to 1/0 for the category value of the nominal scale data, and converting the order/interval/ratio scale data to 1/0 for the category value using the correspondence table, the application form Convert the entire information into a 1/0 table for category values. Furthermore, cluster analysis is performed based on the distance between the applications in the converted 1/0 table to calculate application clusters. Based on the calculated number of records belonging to each cluster and the ratio of NG examination results to the number of belonging records, a group of application forms with a high possibility of being an error case is extracted. Verification is performed for some of the application forms extracted above. Therefore, when extracting and verifying some application forms from all the application forms, the detection rate of business error cases can be increased.

1000 business system 10 business support server 20 business server 30 business client terminal 5 network 101 storage unit 11 application form information data 12 category definition 12 of nominal scale data
13 Category definition of order, interval, ratio scale data 14 1/0 data based on category definition 15 Application form-cluster correspondence table 16 Cluster information 17 Main inter-cluster distance information 1A 1/0 data creation unit 1B Cluster analysis unit 1C Business Out-of-work case extraction unit (out-of-work case extraction unit)
1D cluster business exception case extraction unit (cluster business exception case extraction unit)
1E inter-cluster information difference extraction unit 20 business server 201 storage unit 202 calculation unit 202

Claims (4)

An out-of-business case extraction support system for extracting out-of-case cases of application examination work,
generating first 1/0 data for the nominal scale data based on the nominal scale data included in the application and the category values corresponding to the nominal scale data; /ratio scale data and category values corresponding to said ordinal/interval/ratio scale data; a 1/0 data creation unit that creates a 1/0 table including the first 1/0 data and the second 1/0 data;
a cluster analysis unit that performs cluster analysis based on the distance between the applications in the generated 1/0 table and calculates clusters of the applications ;
When the cluster analysis unit counts the number of records belonging to each cluster and the number of records for which the examination result was rejected for each cluster of the calculated application forms and displays them on the display unit,
The 1/0 data creation unit generates the 1/0 table for each of the first application group and the second application group,
The cluster analysis unit counts the number of records belonging to each cluster and the number of records for which the screening result was negative for each of the first application group and the second application group, and counts the records belonging to the cluster. displaying on the display unit the difference between the first application form group and the second application form group in terms of the number and the number of records for which the examination result was rejected;
An out-of-business case extraction support system characterized by: The business exception case extraction support system according to claim 1 ,
The cluster analysis unit determines the cluster to which the input new application record belongs, and displays the number of records belonging to the corresponding cluster and the number of records for which the screening result was negative on the display unit.
An out-of-business case extraction support system characterized by: An out-of-service case extraction support method for extracting out-of-service cases of application examination work, comprising:
a 1/0 data generating unit generating first 1/0 data for the nominal scale data based on the nominal scale data included in the application form and the category values corresponding to the nominal scale data;
The 1/0 data creation unit, based on the order/interval/ratio scale data included in the application form and the category value corresponding to the order/interval/ratio scale data, for the order/interval/ratio scale data generating the second 1/0 data of
The 1/0 data creation unit generates a 1/0 table including the first 1/0 data and the second 1/0 data for the application form,
When the cluster analysis unit performs cluster analysis based on the distance between the applications in the generated 1/0 table and calculates the cluster of the applications ,
When the cluster analysis unit counts the number of records belonging to each cluster and the number of records for which the examination result was rejected for each cluster of the calculated application forms and displays them on the display unit,
The 1/0 data creation unit generates the 1/0 table for each of the first application group and the second application group,
The cluster analysis unit counts the number of records belonging to each cluster and the number of records for which the screening result was negative for each of the first application group and the second application group, and counts the records belonging to the cluster. displaying on the display unit the difference between the first application form group and the second application form group in terms of the number and the number of records for which the examination result was rejected;
An out-of-work case extraction support method characterized by: The work outlier case extraction support method according to claim 3 ,
The cluster analysis unit determines the cluster to which the input new application record belongs, and displays the number of records belonging to the corresponding cluster and the number of records for which the screening result was negative on the display unit.
An out-of-work case extraction support method characterized by:

Images