JP7617792B2 - Fault classification inference correction device - Google Patents

Description

This disclosure relates to a fault classification inference correction device used when analyzing fault information that occurs during software development.

In software development, in order to reduce development costs, it is necessary to reduce the number of defects that occur during development and after the product (the product equipped with the developed software) is shipped. One possible way to reduce the number of defects is to identify development phases (requirements definition, basic design, detailed design, implementation, etc.) that are likely to cause defects, and to increase the amount of resources allocated to those development phases.

When using a development management system for software development, when a problem occurs, the problem information about the problem and a record of the activities performed to deal with the problem are managed in the development management system in the form of a problem ticket. With the goal of reducing software development costs, the information contained in multiple problem tickets may be aggregated in order to review past problem tickets and identify the development phase in which the problem originated.

A problem ticket contains information about the period after the problem was recognized, but in many cases it does not contain information about the cause of the problem, such as how the defect was embedded, which is information that exists before the problem was recognized. When identifying the development phase in which the problem originated, the work required to manually check the activity records contained in the problem ticket and various pieces of information referenced from those activity records is very time-consuming. Therefore, there is a need for a mechanism that can automatically infer the development phase in which the problem originated from the problem ticket.

Here, we consider building the above system by using AI (Artificial Intelligence) technology to train a fault information database that stores past fault tickets, and then applying a classification algorithm based on the learning results.

In order to improve the accuracy of AI technology, a technique has been disclosed in which corrections are made to data at the data preparation stage before learning, based on similarity metrics that evaluate the similarity of the data, to make classification clearer (see, for example, Patent Document 1).

Patent No. 6568935

However, conventional methods have not provided a method for calculating the specific evaluation parameters required to classify problem information, such as the various situations at the development site, such as the amount of discussions that have been held by stakeholders in different positions in the development management system, or the extent to which problem information has been referenced when accumulating know-how at the development site. As a result, conventional methods have not been able to accurately infer the development phase in which a problem occurs, and there is room for improvement.

The present disclosure has been made to solve such problems, and aims to provide a fault classification inference correction device that can contribute to accurate inference of the development phase where the fault originates.

To solve the above problems, the fault classification inference correction device according to the present disclosure includes a first evaluation unit that evaluates the relevance between a fault ticket including information about a fault that occurred in software development and an element parameter that is a parameter related to the fault ticket, a second evaluation unit that evaluates the relevance between the fault ticket and a development phase based on a first evaluation value that is the evaluation result of the first evaluation unit and a second evaluation value that indicates the relevance between the development phase of software development and the element parameter, and a correction unit that applies a correction based on a third evaluation value that is the evaluation result of the second evaluation unit to the fault classification inference unit that infers the development phase in which the fault is caused for an input fault ticket that is the subject of classification based on fault information with classification information in which information about the development phase in which the fault is caused is added to the fault ticket.

This disclosure can contribute to accurate inference of the development phase in which a fault may occur.

1 is a block diagram showing a configuration of a fault classification inference correction device according to an embodiment. FIG. 2 is a diagram illustrating an example of a failure information DB according to the embodiment. FIG. 11 is a diagram showing an example of element parameters according to the embodiment. FIG. 2 is a diagram illustrating an example of a configuration of a trouble ticket-element parameter relevance evaluation unit according to an embodiment. FIG. 13 is a diagram showing an example of a trouble ticket-element parameter relevance evaluation value according to an embodiment. FIG. 13 is a diagram showing an example of a trouble ticket-element parameter relevance evaluation value according to an embodiment. FIG. 13 is a diagram showing an example of a development phase-element parameter relevance evaluation value according to an embodiment. FIG. 13 is a diagram showing an example of a development phase-element parameter relevance evaluation value according to an embodiment. FIG. 13 is a diagram showing an example of calculation of a trouble ticket-development phase relevance evaluation value according to an embodiment. FIG. 13 is a diagram showing an example of a trouble ticket-development phase relevance evaluation value according to an embodiment. FIG. 11 is a diagram illustrating an example of a classification information-attached failure information DB according to the embodiment. FIG. 13 is a diagram showing an example of an input failure ticket according to an embodiment. FIG. 2 is a diagram illustrating an example of a configuration of a fault classification inference correction unit according to an embodiment. 2 is a block diagram showing an example of a configuration of a fault classification inference unit according to an embodiment. FIG. FIG. 13 is a diagram showing an example of a fault classification inference result according to an embodiment. FIG. 4 is a diagram showing an example of teacher data according to the embodiment. FIG. 11 is a diagram showing an example of input data according to the embodiment. A figure showing an example of added teacher data according to an embodiment. FIG. 13 is a diagram showing an example of input data after addition according to an embodiment. FIG. 11 is a diagram showing an example of an inference result according to the embodiment. A figure showing an example of calculation of a corrected inference result according to an embodiment. A figure showing an example of a corrected inference result according to an embodiment. 1 is a block diagram showing an example of a hardware configuration of a fault classification inference correction device according to an embodiment. 1 is a block diagram showing an example of a hardware configuration of a fault classification inference correction device according to an embodiment.

<Embodiment>
<Configuration>
FIG. 1 is a block diagram showing a configuration of a fault classification inference correction device 1 according to an embodiment.

The fault classification inference correction device 1 includes a fault ticket-element parameter relevance evaluation unit 4 (first evaluation unit), a fault ticket-development phase relevance evaluation unit 7 (second evaluation unit), and a fault classification inference correction unit 11. The fault classification inference correction device 1 receives as input a fault ticket and element parameters 3 stored in a fault information DB (Database: the same applies below) 2, and performs corrections on the fault classification inference unit 12.

Figure 2 shows an example of fault information DB2.

The problem information DB2 stores information on at least one problem ticket 201. The problem ticket 201 includes problem information 202, access history 203, person in charge history 204, comment history 205, and citation history 206. In other words, the problem ticket 201 includes information on a problem that occurred during software development. Note that in FIG. 2, a problem ticket Foo is shown as an example of the problem ticket 201.

The trouble information 202 includes the trouble response period, the trouble content, the trouble response content, and the status. The access history 203 includes the date and time when the trouble ticket was accessed, the accessor who accessed the trouble ticket, and the location from which the accessor accessed the trouble ticket immediately after accessing the trouble ticket.

The person in charge history 204 includes the date and time when the information of the person in charge related to the trouble ticket 201 was changed, the type of event that was changed, the person in charge before the change (person in charge (From)), and the person in charge after the change (person in charge (To)). In other words, the person in charge history 204 is the history of the person in charge who was in charge of the trouble ticket 201.

The comment history 205 includes the date and time when a comment was posted on the problem ticket 201, the commenter, and the comment content. The citation history 206 includes the date and time when the problem ticket 201 was cited from another article in the development management system, the citer, and the location of the article that cited the problem ticket (citation location).

Figure 3 shows an example of element parameter 3.

The element parameters 3 are parameters related to the trouble ticket 201 stored in the trouble information DB 2. Examples of the element parameters 3 include parameters such as the person involved 301 or article 302 related to the trouble ticket 201.

Figure 4 shows an example of the configuration of the trouble ticket-element parameter relevance evaluation unit 4.

The trouble ticket-element parameter relevance evaluation unit 4 includes an access count extraction unit 401, a comment count extraction unit 402, a related person URL attachment count extraction unit 403, a person in charge period extraction unit 404, an article URL attachment count extraction unit 405, and a immediately followed access article extraction unit 406. The trouble ticket-element parameter relevance evaluation unit 4 may be configured to include at least one of these extraction units.

The access count extraction unit 401 counts the number of times an accesser has accessed the trouble ticket 201 from the information in the access history 203 in the trouble ticket 201. If there are multiple accessers, the number of times is counted for each accesser. An accesser corresponds to the related party, which is the element parameter 3.

The comment count extraction unit 402 counts the number of times a commenter has posted a comment on the problem ticket 201 from the information in the comment history 205 of the problem ticket 201. If there are multiple commenters, the count is calculated for each commenter. A commenter corresponds to a related party, which is element parameter 3.

The related party URL attachment count extraction unit 403 counts the number of times that a quoter has attached the URL of a problem ticket to an article (the number of times that a related party has cited the problem ticket) from the information in the citation history 206 of the problem ticket 201. If there are multiple quoters, the count is tallied for each quoter. The quoters and articles correspond to the related party and article, which are element parameters 3, respectively.

The person in charge period extraction unit 404 tally up the period during which a person was in charge of the problem ticket 201 from the information in the person in charge history 204 of the problem ticket 201. If there are multiple people in charge, the period is tallied for each person in charge.

The article URL attachment count extraction unit 405 counts the number of times the URL of a trouble ticket has been attached to an article (the number of times the trouble ticket has been cited in an article) from the information in the citation history 206 of the trouble ticket 201. If there are multiple articles, the count is calculated for each article.

The immediately-accessed article extraction unit 406 counts the number of times an accesser accessed an article immediately after a trouble ticket from the information in the access history 203 in the trouble ticket 201. Note that if there are multiple articles, the number of times is counted for each article.

Figures 5 and 6 are diagrams showing an example of a problem ticket-element parameter relevance evaluation value 5 (first evaluation value). The problem ticket-element parameter relevance evaluation value 5 is the evaluation result of the problem ticket-element parameter relevance evaluation unit 4, and includes a related person-problem ticket relevance evaluation value 501 (Figure 5) and an article-problem ticket relevance evaluation value 502 (Figure 6). In Figures 5 and 6, the problem ticket 201 is problem ticket Foo.

As shown in FIG. 5, the related party-trouble ticket relevance evaluation value 501 indicates the relevance between the related party and the trouble ticket 201. The evaluation parameters 503 include access 504, comments 505, URL attachment to other articles 506, and responsible party period 507.

Access 504 is an evaluation value calculated by the access count extraction unit 401. In FIG. 5, it is shown that Mr. A, a related person (accessor), accessed the trouble ticket Foo eight times.

Comment 505 is an evaluation value calculated by the comment count extraction unit 402. In FIG. 5, it is shown that Mr. B, a related party (commenter), has posted six comments on the problem ticket Foo.

The URL pasted in another article 506 is an evaluation value calculated by the related party URL pasting count extraction unit 403. In FIG. 5, it is shown that Mr. C, a related party (quoter), pasted the URL of the trouble ticket Foo once in another article.

The person in charge period 507 is an evaluation value calculated by the person in charge period extraction unit 404. In FIG. 5, it is shown that the period during which the person in charge A, who is the related party (person in charge), was in charge of the trouble ticket 201 was one day.

As shown in FIG. 6, the article-trouble ticket relevance evaluation value 502 indicates the relevance between the article and the trouble ticket 201. The evaluation parameters 503 include attaching the URL of the trouble ticket to the article 508, and accessing the article immediately after the trouble ticket 509.

The URL of the trouble ticket is attached to the article 508, which is an evaluation value calculated by the article URL attachment count extraction unit 405. In FIG. 6, it shows that the URL of the trouble ticket Foo was attached to the article Hoge once.

The article accessed immediately after a trouble ticket 509 is an evaluation value calculated by the immediately accessed article extraction unit 406. In FIG. 6, it is shown that the number of times that an accessor accessed article Fuga immediately after a trouble ticket Foo was 7.

Figures 7 and 8 are diagrams showing an example of a development phase-element parameter relevance evaluation value 6 (second evaluation value). The development phase-element parameter relevance evaluation value 6 includes a stakeholder-development phase relevance evaluation value 601 (Figure 7) and an article-development phase relevance evaluation value 602 (Figure 8). In Figures 7 and 8, phase refers to the development phase.

The stakeholder-development phase relevance evaluation value 601 is derived based on the structure of the development project, etc. In FIG. 7, the stakeholder-development phase relevance evaluation value 601 for stakeholder A is "0.7" for phase 0, "0.2" for phase 1, and "0.1" for phase 2.

The article-development phase relevance evaluation value 602 is derived from the posting date and time of the article, etc. In FIG. 8, the article-development phase relevance evaluation value 602 for the article Hoge is "0.6" for phase 0, "0.1" for phase 1, and "0.2" for phase 2.

Figure 9 shows an example of the calculation of the problem ticket-development phase relevance evaluation value 8 (third evaluation value).

As shown in FIG. 9, the trouble ticket-development phase relevance evaluation unit 7 calculates a trouble ticket-development phase relevance evaluation value 8 based on the trouble ticket-element parameter relevance evaluation value 5 and the development phase-element parameter relevance evaluation value 6.

Figure 10 shows an example of a problem ticket-development phase relevance evaluation value 8.

In FIG. 10, the problem ticket-development phase relevance evaluation value 8 for problem ticket Foo is "0.24" for phase 0, "0.77" for phase 1, and "0.58" for phase 2.

Figure 11 shows an example of fault information DB9 with classification information.

The fault information DB9 with classification information stores the fault tickets 201 in the fault information DB2 by adding information on the development phase 901 in which the fault originated, based on manual confirmation or the like.

Figure 12 shows an example of an input error ticket 10.

The input problem ticket 10 is a problem ticket that is to be classified when identifying the development phase in which the problem originates. The input problem ticket 10 corresponds to one of the problem tickets 201 stored in the problem information DB 2 for which the development phase in which the problem originates has not been identified.

Figure 13 shows an example of the configuration of the fault classification inference correction unit 11.

The fault classification inference correction unit 11 includes a feature item addition unit 1101 and an inference result correction unit 1102. The fault classification inference correction unit 11 may be configured to include only either the feature item addition unit 1101 or the inference result correction unit 1102. Details of the feature item addition unit 1101 and the inference result correction unit 1102 will be described later.

Figure 14 is a block diagram showing an example of the configuration of the fault classification inference unit 12.

The fault classification inference unit 12 includes preprocessing 1201, learning 1206, and inference 1208. These will be described in detail later.

Figure 15 shows an example of a fault classification inference result 13.

The fault classification inference result 13 is the development phase in which the fault for the input fault ticket 10 is caused. In FIG. 15, the inference result for the input fault ticket Bar indicates that it is phase 1.

Figure 16 shows an example of teacher data 1202.

The teacher data 1202 consists of a set of feature items and features, and is obtained by applying a general feature extraction algorithm to the classification information-attached fault information DB 9 as preprocessing 1201. In FIG. 16, the feature item is the "occurrence frequency of the word 'current'" and the feature amount is "0.6".

Figure 17 shows an example of input data 1203.

The input data 1203 consists of a set of feature items and features, and is obtained by applying a general feature extraction algorithm to the input trouble ticket 10 as preprocessing 1201. In FIG. 17, the feature item is the "occurrence frequency of the word 'electric current'" and the feature is "0.2".

Figure 18 shows an example of added teacher data 1204.

The added teacher data 1204 is obtained by adding information on the trouble ticket-development phase relevance evaluation value 8 to the teacher data 1202 as a pair of feature items and features by the feature item adding unit 1101.

Figure 19 shows an example of input data 1205 after addition.

The added input data 1205 is the input data 1203 to which the feature item adding unit 1101 has added information on the trouble ticket-development phase relevance evaluation value 8 as a pair of feature items and features.

Figure 20 shows an example of an inference result 1209.

The inference result 1209 shows the degree of conformance of the development phase that is the cause of the fault for the input fault ticket 10, obtained by the inference 1208. In FIG. 20, the degree of conformance of the development phase that is the cause of the fault for the fault ticket Bar is "0.67" for phase 0, "0.72" for phase 1, and "0.18" for phase 2.

Figure 21 shows an example of the calculation of the corrected inference result 1210.

As shown in FIG. 21, the inference result correction unit 1102 corrects the inference result 1209 (the degree of conformance of the development phase in which the fault originates) based on the inference result 1209 and the fault ticket-development phase relevance evaluation value 8.

Figure 22 shows an example of a corrected inference result 1210.

The corrected inference result 1210 is calculated by the inference result correction unit 1102 correcting the inference result 1209. In FIG. 22, the conformance degree of the development phase that causes the fault for the fault ticket Bar is "0.72" for phase 0, "0.63" for phase 1, and "0.28" for phase 2.

<Algorithm of the fault classification inference unit 12>
The algorithm of the fault classification inference unit 12 will be described in detail with reference to FIG.

First, in preprocessing 1201, the trouble tickets stored in the classification information-attached trouble information DB 9 are converted into teacher data 1202, and the input trouble tickets 10 are converted into input data 1203.

Next, when the feature item adding unit 1101 is used, the trouble ticket-development phase relevance evaluation value 8 is added to the teacher data 1202 to obtain the added teacher data 1204. Also, the trouble ticket-development phase relevance evaluation value 8 is added to the input data 1203 to obtain the added input data 1205. On the other hand, when the feature item adding unit 1101 is not used, the teacher data 1202 and the input data 1203 become the added teacher data 1204 and the added input data 1205 as they are.

Next, learning 1206 is performed on the added training data 1204 to generate a classifier 120 that links the problem ticket with the development phase in which the problem occurred.

Next, inference 1208 is performed on the added input data 1205 by the classifier 1207, and an inference result 1209 indicating the degree of conformance of the development phase in which the fault is caused is obtained.

Next, when the inference result correction unit 1102 is used, a specific calculation formula (for example, see FIG. 21) is applied to the inference result 1209 and the problem ticket-development phase relevance evaluation value 8 to calculate the corrected inference result 1210. On the other hand, when the inference result correction unit 1102 is not used, the inference result 1209 becomes the corrected inference result 1210 as it is.

Finally, the development phase that is most likely to match in the corrected inference result 1210 (the development phase that is most likely to be the cause of the fault in the input fault ticket 10) is set as the fault classification inference result 13.

＜Effects＞
According to the present disclosure, when inference is made for determining the development phase in which a fault is caused for a fault ticket 201 stored in the fault information DB 2 in past development, a more accurate inference result can be obtained by correcting at least one of the pre-learning teacher data 1202 (pre-learning data) and the inference result 1209 according to various situations such as people involved in the development site and other articles on the development system. In other words, the fault classification inference correction device 1 can contribute to accurate inference of the development phase in which a fault is caused.

<Hardware Configuration>
The functions of the fault ticket-element parameter relevance evaluation unit 4, the fault ticket-development phase relevance evaluation unit 7, and the fault classification inference correction unit 11 in the fault classification inference correction device 1 described above are realized by a processing circuit. That is, the fault classification inference correction device 1 includes a processing circuit for evaluating the relevance between the fault ticket and the element parameter, evaluating the relevance between the fault ticket and the development phase, and inferring the development phase in which the fault is caused for the fault ticket. The processing circuit may be dedicated hardware, or may be a processor (also called a CPU, central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, or DSP (Digital Signal Processor)) that executes a program stored in a memory.

When the processing circuit is a dedicated hardware, as shown in FIG. 23, the processing circuit 14 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination of these. Each function of the trouble ticket-element parameter relevance evaluation unit 4, the trouble ticket-development phase relevance evaluation unit 7, and the trouble classification inference correction unit 11 may be realized by the processing circuit 14, or each function may be realized together by a single processing circuit 14.

When the processing circuit 14 is the processor 15 shown in FIG. 24, the functions of the trouble ticket-element parameter relevance evaluation unit 4, the trouble ticket-development phase relevance evaluation unit 7, and the trouble classification inference correction unit 11 are realized by software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in the memory 16. The processor 15 realizes each function by reading and executing the program recorded in the memory 16. That is, the trouble classification inference correction device 1 has a memory 16 for storing a program that results in the execution of a step of evaluating the relevance between the trouble ticket and the element parameter, a step of evaluating the relevance between the trouble ticket and the development phase, and a step of inferring the development phase in which the trouble is caused for the trouble ticket. It can also be said that these programs cause a computer to execute the procedures or methods of the trouble ticket-element parameter relevance evaluation unit 4, the trouble ticket-development phase relevance evaluation unit 7, and the trouble classification inference correction unit 11. Here, memory may be, for example, non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), magnetic disk, flexible disk, optical disk, compact disk, DVD (Digital Versatile Disc), or any storage medium that will be used in the future.

In addition, with regard to the functions of the trouble ticket-element parameter relevance evaluation unit 4, the trouble ticket-development phase relevance evaluation unit 7, and the trouble classification inference correction unit 11, some functions may be realized by dedicated hardware, and other functions may be realized by software or firmware.

In this way, the processing circuitry can realize each of the above-mentioned functions through hardware, software, firmware, or a combination of these.

Note that within the scope of this disclosure, the embodiments may be modified or omitted as appropriate.

1 Trouble classification inference correction device, 2 Trouble information DB, 3 Element parameters, 4 Trouble ticket-element parameter relevance evaluation unit, 5 Trouble ticket-element parameter relevance evaluation value, 6 Development phase-element parameter relevance evaluation value, 7 Trouble ticket-development phase relevance evaluation unit, 8 Trouble ticket-development phase relevance evaluation value, 9 Trouble information DB with classification information, 10 Input trouble ticket, 11 Trouble classification inference correction unit, 12 Trouble classification inference unit, 13 Trouble classification inference result, 14 Processing circuit, 15 Processor, 16 Memory, 201 Trouble ticket, 202 Trouble information, 203 Access history, 204 Person in charge history, 205 Comment history, 206 Citation history, 301 Related person, 302 Article, 401 Access count extraction unit, 402 Comment count extraction unit, 403 Related person URL attachment count extraction unit, 404 Person in charge period extraction unit, 405 Article URL attachment count extraction unit, 406 Immediately accessed article extraction unit, 501 Related party-trouble ticket relevance, 502 Article-trouble ticket relevance, 503 Evaluation parameters, 504 Access, 505 Comment, 506 Attach URL to other article, 507 Person in charge period, 508 Attach URL of trouble ticket to article, 509 Access article immediately after trouble ticket, 601 Related party-development phase relevance, 602 Article-development phase relevance, 603 Development phase, 901 Development phase where trouble is caused, 1101 Feature item addition unit, 1102 Inference result correction unit, 1201 Preprocessing, 1202 Teacher data, 1203 Input data, 1204 Added teacher data, 1205 Added input data, 1206 Learning, 1207 Classifier, 1208 Inference, 1209 Inference result, 1210 Corrected inference result.

Claims (9)

a first evaluation unit that evaluates a relevance between a trouble ticket including information about a trouble that has occurred in software development and an element parameter that is a parameter related to the trouble ticket;
a second evaluation unit that evaluates a relevance between the trouble ticket and the development phase based on a first evaluation value that is an evaluation result of the first evaluation unit and a second evaluation value that indicates a relevance between the development phase of the software development and the element parameter;
a correction unit that applies a correction based on a third evaluation value, which is an evaluation result of the second evaluation unit, to a fault classification inference unit that infers a development phase in which a fault has occurred for an input fault ticket that is a target for classification, based on fault information with classification information to which information on a development phase in which a fault has occurred for the fault ticket has been added;
A fault classification inference correction device comprising: The element parameters include a party associated with the trouble ticket;
the trouble ticket includes a history of the parties' access to the trouble ticket;
The fault classification inference correction device according to claim 1 , wherein the first evaluation unit counts the number of times the concerned person has accessed the fault ticket. The element parameters include a party associated with the trouble ticket;
The trouble ticket includes a history of comments posted by the related party to the trouble ticket;
The fault classification inference correction device according to claim 1 , wherein the first evaluation unit counts the number of times the related person posts a comment on the fault ticket. The element parameters include a party associated with the trouble ticket;
the trouble ticket includes a history of when the participant has cited the trouble ticket;
The fault classification inference correction device according to claim 1 , wherein the first evaluation unit counts the number of times the related person has cited the fault ticket. The element parameters include a party associated with the trouble ticket;
The trouble ticket includes a history of the person involved being an assignee of the trouble ticket;
The fault classification inference correction device according to claim 1 , wherein the first evaluation unit counts up a period during which the person involved was in charge of the fault ticket. The element parameters include an article associated with the trouble ticket;
The trouble ticket includes a history of when the trouble ticket was cited in the article;
The fault classification inference correction device according to claim 1 , wherein the first evaluation unit counts the number of times the fault ticket is cited in the article. The element parameters include an article associated with the trouble ticket;
the trouble ticket includes a history of accesses to the article immediately following the trouble ticket;
The fault classification inference correction device according to claim 1 , wherein the first evaluation unit counts the number of times the article is accessed immediately after the trouble ticket. The correction unit corrects pre- learning data used in the fault classification inference unit ,
The fault classification inference correction device according to claim 1 , wherein the fault classification inference unit performs learning on the pre-learning data corrected by the correction unit . The fault classification inference correction device according to claim 1, wherein the correction unit corrects the inference result inferred by the fault classification inference unit.

Images