JPH11219297A - Diagnostic type problem solving device and program recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To lower a cost and to provide a solution easily understandable to a user in a diagnostic type problem solving device for obtaining the solution to a problem at present. SOLUTION: This device is provided with an example data base 10 for managing examples composed of the pair data of the problem generated in the past and the solution to it, a similarity degree calculation means 12 for calculating a similarity degree between the problem at present and the problem provided in the example stored in the example data base 10, a selection means 13 for selecting the plural examples in the descending order of the similarity degree from the examples stored in the example data base 10, a reliability calculation means 14 for calculating the reliability of the solutions provided in the examples from the similarity degrees of the solutions provided in the examples selected by the selection means 13 and the problems provided in the examples and an output means 17 for outputting the solution to the problem at present based on the reliability calculated by the reliability calculation means 14.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a diagnostic problem solving apparatus for obtaining a solution to a current problem, and a program recording medium on which a program used for realizing the diagnosis is recorded. The present invention relates to a diagnostic problem solving apparatus capable of obtaining a solution that is easy for a user to understand, and a program recording medium on which a program used for realizing the diagnostic problem solving apparatus is recorded.

[0002] In various fields, failure diagnosis problems arise.
For example, printed boards are tested for abnormalities before being shipped. If an error occurs on the printed board, it is analyzed which part of the printed board is causing the error and at which position, and after replacing the part that seems to be abnormal, there is no abnormal part again A test is performed.

At this time, by inputting a symptom relating to an abnormality of the printed board, if there is a diagnosis type problem solving device for diagnosing an abnormal part or an abnormal part, the cost for analyzing and replacing the abnormality can be greatly reduced. become.

[0004] Against this background, there is a need for a diagnostic-type problem solving device that solves a problem currently occurring or assists a user in solving a problem. The solution device needs to be realized at low cost in order to enhance the practicality, and it is necessary to be able to obtain a solution that is easy for the user to understand.

[0005]

2. Description of the Related Art In the prior art, diagnosis-type problem solving has been realized by an expert system.

In this diagnostic type expert system,
The expert's knowledge of problem solving is expressed in the form of rules, and these rules are used to derive a solution to the current problem.

Here, in general, knowledge acquired from experts is stored in a rule database in the form of if-then rules, and when a current problem is input, a rule matching this problem is obtained. Is used to derive a solution to the current problem.

As described above, conventionally, diagnosis-type problem solving has been realized by a diagnosis-type expert system based on knowledge acquisition from specialists.

[0009]

However, in the diagnostic expert system, expert knowledge is stored in a rule database in the form of rules, and the rules stored in the rule database are used to solve the current problem. Deriving the solution had two major challenges.

[0010] The first problem is that the cost required to construct a rule database is very large. That is, acquiring knowledge from a specialist in the form of rules is very difficult and requires a large cost.

The second problem is that it is difficult to explain the derivation of the solution in a format that is easy for the user to understand. In other words, the explanation of the derivation of the solution is performed by displaying the applied rules. However, if the number of displayed rules is enormous, it becomes difficult for the user to understand the explanation.

The present invention has been made in view of the above circumstances, and provides a new diagnostic type problem solving apparatus which can be realized at low cost and can obtain a solution which is easy for a user to understand. It is an object of the present invention to provide a new program recording medium on which a program used for realizing the program is recorded.

[0013]

FIG. 1 shows the principle configuration of the present invention. In the figure, reference numeral 1 denotes a diagnostic problem solving apparatus equipped with the present invention, which obtains and outputs a solution to a current problem when given.

The diagnostic problem solving apparatus 1 of the present invention comprises a case database 10, an input unit 11, and a similarity calculating unit 1.
2, a selecting unit 13, a reliability calculating unit 14, a threshold learning unit 15, a case number learning unit 16, and an output unit 17.

The case database 10 manages cases made up of paired data of a problem that has occurred in the past and a solution to the problem. The input means 11 inputs a current question. The similarity calculating unit 12 calculates the similarity between the current problem input by the input unit 11 and the problem of the case stored in the case database 10.

The selecting means 13 selects a plurality of cases from the cases stored in the case database 10 in descending order of similarity calculated by the similarity calculating means 12. The reliability calculation means 14 calculates the reliability of the solutions of the cases based on the solutions of the cases selected by the selection means 13 and the similarities of the problems of the cases calculated by the similarity calculation means 12. Is calculated.

The threshold learning means 15 assumes a plurality of thresholds when the selecting means 13 selects a case showing a similarity greater than a prescribed threshold, and for each of those thresholds, the case database 10 By evaluating the correct answer rate using the case stored in, the threshold used for the selection process of the selecting unit 13 is obtained from among the thresholds.

The case number learning means 16 assumes a plurality of cases when the selecting means 13 selects a case of a prescribed number of cases, and stores the plurality of cases in the case database 10 for each of those cases. By evaluating the correct answer rate using the stored cases, the number of cases used for the selection process of the selecting unit 13 is obtained from the number of cases.

The output means 17 outputs a solution to the current problem based on the reliability calculated by the reliability calculation means 14. Here, the functions other than the case database 10 of the diagnostic problem solving apparatus 1 of the present invention are specifically realized by a program, and this program is stored on a floppy disk or the like, or stored on a disk such as a server. The present invention is realized by being installed in the diagnostic type problem solving apparatus 1 and operating on the memory.

In the diagnostic problem solving apparatus 1 of the present invention configured as described above, when a current problem is input, the similarity calculating means 12 stores the input current problem and the current problem in the case database 10. The similarity with the problem of the past case is calculated, and the selection unit 13 receives the similarity from the past case stored in the case database 10 based on the threshold value learned by the threshold learning unit 15. A plurality of cases are selected in descending order of the calculated similarity by, for example, selecting cases exhibiting a large similarity or selecting cases corresponding to the number of cases learned by the case number learning means 16.

In response to the case selected by the selection unit 13, the reliability calculation unit 14 calculates the solution of the selected case and the similarity calculated by the similarity calculation unit 11 for the problem of the case. From the above, the reliability which shows the possibility that the solution of the selected case is the solution of the current problem is calculated, and the output means 17 receives the calculated reliability based on the calculated reliability. Output the solution.

As described above, in the diagnostic problem solving apparatus 1 of the present invention, similar cases in the past are specified, and the reliability indicating that the solution of those similar cases is a solution to the current problem is calculated. In this configuration, a solution to the current problem is obtained based on its reliability, so that it is possible to obtain a solution to the current problem without acquiring knowledge from experts. Accordingly, the solution can be realized at low cost and a solution that can be easily understood by the user can be obtained.

[0023]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail according to embodiments. FIG. 2 illustrates an embodiment of the present invention.

In the figure, reference numeral 20 denotes a computer device equipped with the present invention. The computer device 20 includes a diagnostic case database 40 for managing cases composed of pairs of past problems and their solutions. When a current problem is given, a solution to the problem is obtained and output using past cases managed by the diagnostic case database 40.

In order to realize the present invention, the computer device 20 is installed via a floppy disk, a line, or the like, and is created by a problem solving program 21 for obtaining a solution to a current problem, and a problem solving program 21. Similarity table 22 / similar case table 23
/ Reliability table 24 and installed via a floppy disk, line, etc.
Learning program 25 for learning the number of cases and threshold value used in 1
And

The diagnostic case database 40 manages cases composed of paired data of a problem that has occurred in the past and a solution to the problem using, for example, a natural language. As shown in FIG. 3, what kind of failure has occurred in the printed circuit board in the past and which part (such as AF2)
2-40) manages whether the failure has occurred. The former is a problem of past cases, and the latter is a solution of past cases. In addition, as auxiliary information, a case name, the name of a person in charge of the failure, the date of occurrence of the failure, special notes on the failure, and the like are managed.

FIGS. 4 and 5 show a problem solving program 21.
1 illustrates an embodiment of a processing flow executed by the embodiment. Next, the present invention will be described in detail according to this processing flow. When the current problem is given from the terminal 30, the problem solving program 21 first, as shown in the processing flow of FIGS. 4 and 5, in step 1 among the cases stored in the diagnostic case database 40. , One unprocessed case is selected, and in the subsequent step 2, it is determined whether or not all cases have been selected. When it is determined that all cases have not been selected, that is, in step 1, When it is determined that one unprocessed case has been selected, the process proceeds to step 3 where the similarity between the current problem given from the terminal 30 and the problem of the selected case is calculated and the similarity is calculated. After storing in the degree table 22, the process returns to the step 1.

This similarity calculation process uses, for example, the weight of a keyword included in the document describing the problem, adds up the weight of the keyword that is commonly included in both problems, and changes it from “0” to “1”. By performing normalization to a value between ".

Here, the weight of a keyword appearing in a document is, for example, TF (Text Frequency) or IDF (In
verse Document Frequency). This TF represents a weight for the frequency of appearance of the keyword in the document, and is set to have a larger value as the frequency of appearance is higher. On the other hand, IDF
Expresses the variance of the appearance of keywords in the document database, for example, IDF _i = log ₂ [(N− _ni ) / _ni ] where N: number of documents in the document database _ni : number of documents in the document database It is calculated according to a calculation formula of the appearance frequency of the keyword i, and is set to have a larger value as the appearance frequency of the keyword in the document database is lower.

In this way, by repeating Step 1 and Step 3, the problem solving program 21
Is, as shown in FIG.
With the cases stored in the diagnostic case database 40 as registration units, the similarity between the current problem given from the terminal 30 and the problems of those cases is registered.

When the registration of the similarity in the similarity table 22 is completed, and it is determined in step 2 that all cases in the diagnostic case database 40 have been selected, step 4 is performed.
Then, from the cases stored in the similarity table 22, the cases of the prescribed number of cases are selected in descending order of the similarity, and are stored in the similar case table 23.

The number of cases used at this time may be set by interacting with the user. However, if the number determined by the learning program 25 described later is used, the number of cases can be set without subjectivity of the user.

Thus, the problem solving program 21
For example, when five cases are set as the number of cases, a similar case table 23 shown in FIG. 7 is created from the similarity table 22 shown in FIG.

After creating the similar case table 23, one solution is selected in step 5 by selecting one unprocessed case from the cases stored in the similar case table 23. In step 6, when it is determined whether all the solutions (cases) have been selected, and when it is determined that all the solutions (cases) have not been selected, that is, in step 5,
When it is determined that one unprocessed solution (case) has been selected, the process proceeds to step 7 and the similar case table 2
3 is extracted, the reliability of the selected solution is calculated from the similarity of the same solutions, and the calculated reliability is stored in the reliability table 24. Then, the process returns to step S5.
Here, for the extracted cases with the same solution,
It will be treated as selected.

The process of calculating the reliability is performed by, for example, specifying the maximum value of the similarity of the same solution and setting the maximum value as the reliability, or setting the maximum value to the maximum value.
(S _i ), the sum of similarities of the same solution is tot (S _i ), and the sum of similarities stored in the similar case table 23 is TOT.
Then, the reliability of the solution S _i is calculated as follows: reliability (S _i ) = max (S _i ) × [max (S _i ) + [1−max (S _i )] × tot (S _i ) / TOT ] According to the calculation formula.

When the former reliability calculation process is followed, the similar case table 23 shown in FIG.
The reliability of “0.89” is calculated as the reliability for the solution of “/ AF2”. Further, when the latter reliability calculation process is followed, the similarity table 23 shown in FIG. 7 shows the reliability of the solution “MB86932 / AF2” as 0.89 × [0.89+ (1−0.89) × 2.07 / 3.24] The reliability of “0.85” is calculated according to 0.85.

In this way, by repeating steps 5 and 7 instead of step 5, the problem solving program 21
Is, as shown in FIG.
The reliability of the solution to the current problem given from the terminal 30 is registered using the solution of the case stored in the similar case table 23 as a registration unit.

When the registration of the reliability in the reliability table 24 is completed and it is determined in step 6 that all the cases in the similar case table 23 have been selected, the process proceeds to step 8 (FIG. 5).
, The solutions stored in the reliability table 24 are sorted according to the order of the reliability, and in a succeeding step 9, the solutions stored in the reliability table 24 are sorted in the order of the reliability. The output is added to the terminal 30.

When the reliability table 24 shown in FIG. 8 is created in this way, the problem solving program 21 determines that the solution to the current problem given from the terminal 30 is “MB86932” having the reliability “0.85”. / AF2 "," CG21045 / AM5 "having reliability" 0.47 ", and" CA20103 / CB4 "having reliability" 0.31 ".

Upon receiving the output of this solution, the user who has input the current problem will find the most likely solution as "MB
86932 / AF2 ", and as a solution likely to follow, there is" CG21045 / AM5 ". As a solution likely to follow," CA20103 / CB2 "
4 ”will be known.

It should be noted that the problem solving program 21 can be configured to output only the solution with the highest reliability. Next, processing executed by the learning program 25 will be described.

In the processing flows of FIGS. 5 and 6, the problem solving program 21 selects a predetermined number of cases from the cases stored in the similarity table 22 in descending order of similarity. The similar case table 23 is created by storing it in the similar case table 23.
The learning program 25 uses the past cases stored in the diagnostic case database 40 to perform a process of determining the number of cases used at this time by a learning process.

FIG. 9 shows an embodiment of a processing flow executed by the learning program 25. Next, the processing executed by the learning program 25 will be described according to this processing flow.

When a learning request for the number of cases is issued from the user via the terminal 30, the learning program 25 firstly sets the number of cases to the minimum value in step 1 as shown in the processing flow of FIG. Set.

Subsequently, in step 2, it is determined whether or not the number of cases has exceeded the maximum value. When it is determined that the number of cases has not exceeded the maximum value, the process proceeds to step 3, where the diagnostic case database 40 is stored. If one unprocessed case is selected from the stored cases, it is determined in step 4 that all cases have been selected, and if it is determined that all cases have been selected, step 5 is performed. Then, the number of cases is incremented by a specified increment, and the process returns to step 2.

On the other hand, when it is determined in step 4 that all cases have not been selected, that is, in step 3
When it is determined that one unprocessed case has been selected, the process proceeds to step 6, where the problem of the selected case is set as the current problem, and the problem solving program 21 is started. Get a solution to the current problem you set.

When started in this way, the problem solving program 21 calculates the similarity between the set problem and the problem of the case remaining in the diagnostic case database 40, and calculates the similarity. Since the similar case table 23 is created according to the value and the set number of cases and the reliability is calculated from the similar case table 23, a solution to the set problem is obtained. By activating the program 21, a solution to the set current problem is obtained.

Subsequently, in step 7, it is determined whether or not the solution obtained in step 6 is a correct answer. When it is determined that the answer is correct, the process proceeds to step 8, where the score is incremented by one. When the process returns to step 3 to determine that the answer is not correct, the process returns to step 3 without incrementing the score.

As described above, the problem solving program 21
Outputs the solution to the set problem in the order of higher reliability, so the learning program 25 determines the solution of the set problem (the solution of the selected case) and the problem solving program 2
When it is determined whether or not the solution obtained in step 6 is a correct answer by determining whether or not the solution having the highest reliability output from 1 is the same, when determining that the solution is correct, The score is incremented by one.

Here, for example, the problem solving program 21
The score is incremented by 3 points when it matches the solution having the first degree of reliability, and the score is incremented by 2 points when it matches the solution having the second degree of reliability.
It is also possible to adopt a method of incrementing the score by one point when the value matches a solution having a degree of reliability.

When it is determined in step 2 that the number of cases exceeds the maximum value when the score for each set number of cases is obtained in this way, the process proceeds to step 9 to indicate the highest score. The number of cases is determined as the number of cases used by the problem solving program 21, and the process ends.

According to the learning process of the learning program 25, for example, when the learning program 25 determines five as the number of cases, as described above, the similarity table 22 shown in FIG. 23 will be created.

In the processing flows shown in FIGS. 5 and 6, the problem solving program 21 selects a predetermined number of cases from the cases stored in the similarity table 22 in descending order of similarity. Although the similar case table 23 is created by storing it in the similar case table 23, a case having a similarity greater than a specified threshold value is selected and stored in the similar case table 23. May be adopted to create the similar case table 23.

In the case of adopting this configuration, the learning program 25 executes the processing flow of FIG. 10 to determine the threshold value used at this time by the learning processing.
That is, when a learning request for a threshold is issued from the user via the terminal 30, the learning program 25 first sets the threshold to the minimum value in step 1 as shown in the processing flow of FIG. 10.

Subsequently, in step 2, it is determined whether or not the threshold value has exceeded the maximum value. When it is determined that the threshold value has not been exceeded, the process proceeds to step 3, where the threshold value is stored in the diagnostic case database 40. One unprocessed case is selected from the cases to be processed, and it is determined in step 4 that all cases have been selected. When it is determined that all cases have been selected, step 5 is performed. Proceeding, the threshold value is incremented by a specified increment, and the process returns to step 2.

On the other hand, when it is determined in step 4 that all cases have not been selected, that is, in step 3
When it is determined that one unprocessed case has been selected, the process proceeds to step 6, where the problem of the selected case is set as the current problem, and the problem solving program 21 is started. Get a solution to the current problem you set.

When started in this manner, the problem solving program 21 calculates the similarity between the set problem and the problem of the case remaining in the diagnostic case database 40, and calculates the similarity. Since the similar case table 23 is created according to the value and the set threshold value and the reliability is calculated from the similar case table 23, the solution to the set problem is obtained. By starting 21, a solution to the set current problem is obtained.

Subsequently, in step 7, it is determined whether or not the solution obtained in step 6 is a correct answer. When it is determined that the answer is correct, the process proceeds to step 8, where the score is incremented by one. When the process returns to step 3 to determine that the answer is not correct, the process returns to step 3 without incrementing the score.

When it is determined in step 2 that the threshold value has exceeded the maximum value when the score for each of the set threshold values is obtained in this manner, the process proceeds to step 9 where the threshold value indicating the highest score is determined. It is determined as the threshold value used by the problem solving program 21 and the process is terminated.

According to the learning process of the learning program 25, for example, the learning program 25 sets the threshold to "0.6".
Is determined from the similarity table 22 shown in FIG.
The similar case table 23 shown in FIG. 11 is created.

Further, it is also possible to request creation of a similar case table 23 that satisfies both conditions of the number of cases and the threshold value. For example, when the learning program 25 determines five as the number of cases and determines “0.5” as the threshold, it is also possible to request creation of the similar case table 23 that satisfies both conditions. In this case, a similar case table 23 shown in FIG. 12 is created from the similarity table 22 shown in FIG.

[0062]

As described above, in the diagnostic problem solving apparatus of the present invention, similar cases in the past are specified, and the reliability of these similar cases is a solution to the current problem. Because the configuration that calculates the degree is adopted and the solution to the current problem is obtained based on the reliability, the solution to the current problem can be obtained without acquiring the knowledge from the experts. Thus, this can be realized at low cost, and a solution that can be easily understood by the user can be obtained.

When adopting this configuration, the quality of the output solution can be improved by using a configuration that avoids using similar cases with low reliability for problem solving by using the number of cases and thresholds. Can be achieved.

Since a configuration having a mechanism for automatically determining the number of cases and the threshold value is employed, the load on the user can be reduced and the quality of the output solution can be improved.

[Brief description of the drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is an embodiment of the present invention.

FIG. 3 is an explanatory diagram of a diagnostic case database.

FIG. 4 is a processing flow executed by a problem solving program.

FIG. 5 is a processing flow executed by a problem solving program.

FIG. 6 is an explanatory diagram of a similarity table.

FIG. 7 is an explanatory diagram of a similar case table.

FIG. 8 is an explanatory diagram of a reliability table.

FIG. 9 is a processing flow executed by a learning program.

FIG. 10 is a processing flow executed by a learning program.

FIG. 11 is an explanatory diagram of a similar case table.

FIG. 12 is an explanatory diagram of a similar case table.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Diagnosis type problem solving device 10 Case database 11 Input means 12 Similarity calculation means 13 Selection means 14 Reliability calculation means 15 Threshold learning means 16 Case number learning means 17 Output means

Claims (5)

[Claims] 1. A diagnostic problem solving apparatus for obtaining a solution to a current problem, comprising: a case database for managing cases composed of pairs of past problems and solutions to the problems; A similarity calculating means for calculating a similarity between the stored case and the problem, and a selecting means for selecting a plurality of cases in descending order of the similarity from the cases stored in the case database; From the solution of the case selected by the selection unit and the similarity of the problem of the case, the reliability calculation unit that calculates the reliability of the solution, and the reliability calculated by the reliability calculation unit Output means for outputting a solution to the current problem based on the diagnostic problem solving device. 2. The diagnostic problem solving apparatus according to claim 1, wherein the selecting means selects a case exhibiting a degree of similarity greater than a prescribed threshold from among the prescribed number of cases. Diagnostic type problem solving device. 3. The diagnostic problem solving apparatus according to claim 2, wherein a plurality of thresholds are assumed, and the correct answer rate is evaluated for each of the thresholds by using a case stored in a case database. A diagnostic problem solving apparatus, comprising: threshold learning means for obtaining a threshold used for selection processing by the selection means from among the thresholds. 4. The diagnostic problem solving apparatus according to claim 2, wherein a plurality of cases are assumed, and a correct answer rate is evaluated for each of the cases by using a case stored in a case database. And a case number learning means for obtaining the number of cases used in the selection processing of the selection means from the number of cases. 5. A program recording medium for recording a program used for realizing a diagnosis-type problem solving apparatus for obtaining a solution to a current problem, comprising: a pair of data of a problem that occurred in the past and a solution to the problem. A similarity calculation process of calculating a similarity between a current problem and a problem of a case stored in the case database, with the case database that manages the case database as an access destination; From the selection process of selecting a plurality of cases in the order of the similarity, the reliability of the solution is calculated from the solution of the case selected by the selection process and the similarity of the problem of the case. A computer that executes a reliability calculation process to perform a process of outputting a solution to a current problem based on the reliability calculated by the reliability calculation process. That ram is recorded, a program recording medium characterized.