Method and system for evaluating and modifying fuzzy knowledge

A fuzzy knowledge evaluation apparatus for use with a conventional fuzzy reasoning system in which a fuzzy reasoning is achieved by use of real data of a plurality of test cases based on fuzzy knowledge including a plurality of fuzzy rules each described in a format having an if part and a then part, and a plurality of membership functions defining meanings of respective propositions described in the if and then parts, said fuzzy knowledge evaluation apparatus comprising:

• fuzzy knowledge base storage means for storing said plurality of fuzzy rules and said plurality of fuzzy membership functions;
• fuzzy reasoning unit means connected to said fuzzy knowledge base storage means for executing the fuzzy reasoning by performing a fuzzy inference;
• evaluating means operatively associated with the fuzzy reasoning unit means for evaluating a result obtained by calculating a plurality of relationships using each of:
• i) results of the fuzzy inference, ii) intermediate results of the inference, and iii) said real data, wherein the plurality of relationships include a relationship between the results of the inference and the real data, a relationship between the intermediate results of the inference and the real data, a relationship between the fuzzy rules, and a relationship between the membership functions and the real data, said evaluating means including:
• means for obtaining, for each of the fuzzy rules, a grade of rule representing an adequacy thereof based on the membership functions associated with the respective if parts of the fuzzy rules;
• means for obtaining representative values of the membership functions related to the respective then parts of the fuzzy rules for all the cases;
• means for obtaining differences between the representative values and the real data;
• and,
• means for producing a table indicating the relationship between the rule and the grade of the rule on the basis of the representative values obtained by said representative values obtaining means and a grade of rule obtained in the course of executing the fuzzy inference;
• and,
• an output unit operatively associated with the fuzzy reasoning unit means for graphically outputting each of the relationships obtained by said evaluating means.

A fuzzy knowledge evaluation apparatus for use with a conventional fuzzy reasoning system in which a fuzzy reasoning is achieved by use of real data of a plurality of test cases based on fuzzy knowledge including a plurality of fuzzy rules each described in a format having an if part and a then part, and a plurality of membership functions defining meanings of respective propositions described in the if and then parts, said fuzzy knowledge evaluation apparatus comprising:

• fuzzy knowledge base storage means for storing said plurality of fuzzy rules and said plurality of fuzzy membership functions;
• fuzzy reasoning unit means connected to said fuzzy knowledge base storage means for executing the fuzzy reasoning by performing a fuzzy inference;
• evaluating means operatively associated with the fuzzy reasoning unit means for evaluating a result obtained by calculating a plurality of relationships using each of:
• i) results of the fuzzy inference, ii) intermediate results of the inference, and iii) said real data, wherein the plurality of relationships include a relationship between the results of the inference and the real data, a relationship between the intermediate results of the inference and the real data, a relationship between the fuzzy rules, and a relationship between the membership functions and the real data, said evaluating means including:
• means for attaining an influence power of each of the fuzzy rules with respect to the fuzzy reasoning results for all the cases;
• means for assigning values of order to the respective fuzzy rules in a sequence of the magnitude of the obtained influence power for all the cases;
• and
• means for producing a table indicating the relationship between the rules on basis of the orders of all the cases in the fuzzy rules;
• and,
• an output unit operatively associated with the fuzzy reasoning unit means for graphically outputting each other relationships obtained by said evaluating means.

A fuzzy knowledge evaluation apparatus for use with a conventional fuzzy reasoning system in which a fuzzy reasoning is achieved by use of real data of a plurality of test cases based on fuzzy knowledge including a plurality of fuzzy rules each described in a format having an if part and a then part, and a plurality of membership functions defining meanings of respective propositions described in the if and then parts, said fuzzy knowledge evaluation apparatus comprising:

• fuzzy knowledge base storage means for storing said plurality of fuzzy rules and said plurality of fuzzy membership functions;
• fuzzy reasoning unit means connected to said fuzzy knowledge base storage means for executing the fuzzy reasoning by performing a fuzzy inference;
• evaluating means operatively associated with the fuzzy reasoning unit means for evaluating a result obtained by calculating a plurality of relationships using each of:
• i) results of the fuzzy inference, ii) intermediate results of the inference, and iii) said real data, wherein the plurality of relationships include a relationship between the results of the inference and the real data, a relationship between the intermediate results of the inference and the real data, a relationship between the fuzzy rules, and a relationship between the membership functions and the real data, said evaluating means including:
• means for superimposing the real data used as input data during performing the fuzzy inference on the membership functions of the if part;
• and
• means for superimposing the reasoning results on the membership functions of the then part, a result of the superimposition by said means indicating a relationship between the membership function and the real data;
• and,
• an output unit operatively associated with the fuzzy reasoning unit means for graphically outputting each of the relationships obtained by said evaluating means.

A fuzzy knowledge evaluation and modification system for use in a fuzzy reasoning system of known type in which a fuzzy reasoning is achieved by use of fuzzy knowledge including i) fuzzy rules described in a format of an if . . . part and a then . . . part, and ii) membership functions defining meanings of respective propositions described in the if . . . and then . . . parts, said fuzzy knowledge evaluation and modification system comprising:

• fuzzy knowledge base means for storing said fuzzy rules and membership functions;
• a fuzzy reasoning unit for performing said fuzzy reasoning;
• time series data storage means for storing actual time series data;
• processing means for processing, based on the actual time series data stored in said time series data storage means and collected from a field selected as an object of the fuzzy reasoning, values of the if . . . parts of the fuzzy rules, thereby attaining a grade related to the if . . . parts;
• situation value attaining means for attaining situation values of the then . . . parts of the fuzzy rules based on the actual time series data;
• situation producing means for producing a distribution of situations including a plurality of pairs each comprising the grade related to the if . . . parts and the situation value of the then . . . parts;
• grade calculating means for calculating either one of an effective grade and an inconformity grade of the fuzzy rules based on the generated distribution of situations and the membership functions of the then . . . parts of the fuzzy rules;
• and
• display means for displaying the calculated grade.

A fuzzy knowledge evaluation method for use in a conventional fuzzy reasoning system of known type in which a fuzzy reasoning is achieved by use of real test data of a plurality of real test cases based on fuzzy knowledge including a plurality of fuzzy rules each described in a format of an if part and a then part and a plurality of membership functions defining meanings of respective propositions described in the if and then parts, said fuzzy knowledge evaluation method comprising the steps of:

• executing the fuzzy reasoning by performing a fuzzy inference in the fuzzy reasoning system using the plurality of fuzzy rules of the fuzzy knowledge to obtain a result;
• obtaining, in an evaluating means of said fuzzy reasoning system, a plurality of relationships using each of:
• i) the result obtained by the fuzzy inference, ii) intermediate results thereof and iii) the real data, wherein the plurality of relationships include a relationship between the results of inference and the real data, a relationship between the intermediate results of the inference and the real data, a relationship between first rules and grades of the first rules, a relationship between said fuzzy rules, and a relationship between first membership functions and the real data;
• graphically outputting each of said plurality of relationships obtained by said evaluating means as a graphic display representing adequacy of said fuzzy knowledge;
• and,
• adjusting said fuzzy knowledge based on said adequacy to obtained adjusted fuzzy knowledge.

A fuzzy knowledge evaluation method for use in a conventional fuzzy reasoning system of known type in which a fuzzy reasoning is achieved by use of real test data of a plurality of real test cases based on fuzzy knowledge including a plurality of fuzzy rules each described in a format of an if part and a then part and a plurality of membership functions defining meanings of respective propositions described in the if and then parts, said fuzzy knowledge evaluation method comprising the steps of:

• executing the fuzzy reasoning by performing a fuzzy inference in the fuzzy reasoning system using the plurality of fuzzy rules of the fuzzy knowledge to obtain a result;
• obtaining, in an evaluating means of said fuzzy reasoning system, a plurality of relationships using each of:
• i) the result obtained by the fuzzy inference, ii) intermediate results thereof and iii) the real data, wherein the plurality of relationships include a relationship between the results of inference and the real data, a relationship between the intermediate results of the inference and the real data, a relationship between first rules and grades of the first rules, a relationship between said fuzzy rules, and a relationship between first membership functions and the real data, said obtaining step including:
• obtaining, for each of the fuzzy rules, a grade of rule representing an adequacy thereof based on the membership functions associated with the respective if parts of the fuzzy rules;
• attaining representative values of the membership functions related to the respective then parts of each of the fuzzy rules;
• obtaining differences between the representative values and the real data;
• and,
• producing a table indicating the relationship between the first rules and the first grades of the first rules on the basis of the representative values obtained by said representative value attaining step and the grade of rule obtained in the executing of the fuzzy inference;
• graphically outputting each of said plurality of relationships obtained by said evaluating means as a graphic display representing adequacy of said fuzzy knowledge;
• and,
• adjusting said fuzzy knowledge based on said adequacy to obtained adjusted fuzzy knowledge.

A fuzzy knowledge evaluation method for use in a conventional fuzzy reasoning system of known type in which a fuzzy reasoning is achieved by use of real test data of a plurality of real test cases based on fuzzy knowledge including a plurality of fuzzy rules each described in a format of an if part and a then part and a plurality of membership functions defining meanings of respective propositions described in the if and then parts, said fuzzy knowledge evaluation method comprising the steps of:

• executing the fuzzy reasoning by performing a fuzzy inference in the fuzzy reasoning system using the plurality of fuzzy rules of the fuzzy knowledge to obtain a result;
• obtaining, in an evaluating means of said fuzzy reasoning system, a plurality of relationships using each of:
• i) the result obtained by the fuzzy inference, ii) intermediate results thereof and iii) the real data, wherein the plurality of relationships include a relationship between the results of inference and the real data, a relationship between the intermediate results of the inference and the real data, a relationship between first rules and grades of the first rules, a relationship between said fuzzy rules, and a relationship between first membership functions and the real data, said obtaining step including:
• attaining an influence power of each of the fuzzy rules with respect to the fuzzy reasoning results for all the cases;
• assigning values of order to the respective fuzzy rules in a sequence of the magnitude of the obtained influence power for all the cases;
• and
• producing a table indicating the relationship between the rules on the basis of the orders of all the cases in the fuzzy rules;
• graphically outputting each of said plurality of relationships obtained by said evaluating means as a graphic display representing adequacy of said fuzzy knowledge;
• and,
• adjusting said fuzzy knowledge based on said adequacy to obtained adjusted fuzzy knowledge.

A fuzzy knowledge evaluation method for use in a conventional fuzzy reasoning system of known type in which a fuzzy reasoning is achieved by use of real test data of a plurality of real test cases based on fuzzy knowledge including a plurality of fuzzy rules each described in a format of an if part and a then part and a plurality of membership functions defining meanings of respective propositions described in the if and then parts, said fuzzy knowledge evaluation method comprising the steps of:

• executing the fuzzy reasoning by performing a fuzzy inference in the fuzzy reasoning system using the plurality of fuzzy rules of the fuzzy knowledge to obtain a result;
• obtaining, in an evaluating means of said fuzzy reasoning system, a plurality of relationships using each of:
• i) the result obtained by the fuzzy inference, ii) intermediate results thereof and iii) the real data, wherein the plurality of relationships include a relationship between the results of inference and the real data, a relationship between the intermediate results of the inference and the real data, a relationship between first rules and grades of the first rules, a relationship between said fuzzy rules, and a relationship between first membership functions and the real data;
• graphically outputting each of said plurality of relationships obtained by said evaluating means as a graphic display representing adequacy of said fuzzy knowledge, said outputting step including:
• superimposing the real data used as input data during performance of the fuzzy inference on the membership functions of the if part;
• and
• superimposing the reasoning results on the membership functions of the then part
• a result of the superimposing steps indicating a relationship between the membership functions and the real data;
• and,
• adjusting said fuzzy knowledge based on said adequacy to obtained adjusted fuzzy knowledge.

A fuzzy knowledge evaluation and modification method performed in a common fuzzy reasoning apparatus of known type in which a fuzzy reasoning is achieved in a host computer by use of fuzzy knowledge including i) fuzzy rules described in a format of an if . . . part and a then . . . part and ii) membership functions defining meanings of respective propositions described in the if . . . and then . . . parts, said fuzzy knowledge evaluation and modification method comprising the steps of:

• processing in said host computer, based on actual time series data from a field selected as an object of the fuzzy reasoning, values of the if . . . parts of the fuzzy rules, thereby attaining a grade related to the if . . . parts;
• attaining in said host computer situation values of the then . . . parts of the fuzzy rules based on the actual time series data;
• producing a distribution of situations including a plurality of pairs each comprising the grade related to the if . . . parts and the situation value of the then . . . parts;
• calculating either one of an effective grade and an inconformity grade of the fuzzy rules based on the generated distribution of situations and the membership functions of the then . . . parts of the fuzzy rules;
• displaying, on a terminal operatively associated with said host computer, the calculated grade;
• and,
• modifying said fuzzy knowledge using a fuzzy rule editor operatively associated with said host computer based on said displayed calculated grade.