Evolutionary adaptation type inference knowledge extracting apparatus capable of being adapted to a change of input/output date and point of sales data analyzing apparatus using the apparatus

• US 5,740,323 A
• Filed: 04/03/1996
• Issued: 04/14/1998
• Est. Priority Date: 04/05/1995
• Status: Expired due to Fees

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Miyoshi, Tsutomu, Nomura, Tatsuya
• Primary Examiner(s): Hafiz, Tariq R.
• Application Number:
  US08/627,111
• Time in Patent Office: 741 Days
• Field of Search: 395/3, 395/10-13, 395/61, 395/51, 395/50, 395/53, 395/900, 364/550
• US Class Current: 706/13
• CPC Class Codes: G06N 5/048 Fuzzy inferencing
• CPC Class Codes: Y10S 706/90 Fuzzy logic

Abstract

There is provided an inference knowledge extracting apparatus capable of being adapted to a change of input/output data. In a fuzzy rule individual group storing section is stored a group of individuals having a gene string associated with a fuzzy rule of a fuzzy rule storing section by a fuzzy rule gene associating section. A fuzzy rule individual selecting section stochastically selects individuals having a small output error with respect to the input/output data based on a calculation result of fitness obtained by an individual fitness calculating section. A fuzzy rule individual gene manipulating section executes a gene manipulating operation on each individual selected by the individual fitness calculating section. The fuzzy rule gene associating section, individual fitness calculating section, fuzzy rule individual selecting section, fuzzy rule individual gene manipulating section and a rule weight deciding section are functioned, thereby executing an evolutionary adaptation operation to extract a fuzzy rule that is evolutionarily adapted to the change of the input/output data.

Claims

1. An evolutionary adaptation type inference knowledge extracting apparatus comprising:

an input/output data storing section which stores input/output data constituted by a combination of input data comprised of a plurality of numeric values and output data comprised of one discrete value or non-numeric value;

a fuzzy rule gene associating section which associates each of a plurality of fuzzy rules stored in a fuzzy rule storing section with an individual comprised of a gene string for expressing a parameter of each fuzzy membership function of the fuzzy rule by a real number value;

a fuzzy rule individual group storing section which stores a group of a plurality of individuals associated with the fuzzy rule having an identical discrete value or non-numeric value in a posterior part thereof by the fuzzy rule gene associating section;

an individual fitness calculating section which operates, based on a fitness of an anterior part of the fuzzy rule associated with an individual by the fuzzy rule gene associating section with respect to all input/output data stored in the input/output data storing section, to calculates a fitness of the individual;

a fuzzy rule individual selecting section which stochastically selects an individual having a high fitness to the set of the input/output data stored in the input/output data storing section based on the fitness of each individual calculated by the individual fitness calculating section;

a fuzzy rule individual gene manipulating section which generates a new individual by subjecting the individual stored in the fuzzy rule individual group storing section to a gene manipulating operation; and

a rule weight deciding section which operates, based on the fitness calculated by the individual fitness calculating section, to normalize the fitness of each individual every group, thereby deciding a weight of the fuzzy rule corresponding to the individual;

whereby each individual fitted to a characteristic of the input/output data stored in the input/output data storing section is selected, and a fuzzy rule is extracted that is evolutionarily adapted to the input/output data by subjecting the group of the selected individuals to an evolutionary adaptation operation.

2. An evolutionary adaptation type inference knowledge extracting apparatus as claimed in claim 1, wherein

the fuzzy rule individual gene manipulating section has a crossing-over gene manipulating means, using as genes real number values based on two corresponding genes of two individuals belonging to an identical group stored in the fuzzy rule individual group storing section, for executing a crossing-over gene manipulating operation for generating a new individual belonging to the group, and replacing all the individuals in the group with the newly generated individual.

3. An evolutionary adaptation type inference knowledge extracting apparatus as claimed in claim 2, wherein

the crossing-over gene manipulating means has a gene generating means for obtaining an average value of corresponding two genes of the two individuals so as to generate a gene of the new individual.

4. An evolutionary adaptation type inference knowledge extracting apparatus as claimed in claim 2, wherein

the crossing-over gene manipulating means has a gene generating means for putting a real number value which is either one of the corresponding two genes of the two individuals close to or apart from an average value of both the genes so as to generate a gene of the new individual.

5. An evolutionary adaptation type inference knowledge extracting apparatus as claimed in claim 1, wherein

the fuzzy rule individual gene manipulating section has a mutational gene manipulating means for executing a mutational gene manipulating operation for replacing an arbitrary gene of each individual stored in the fuzzy rule individual group storing section with another real number value.

6. An evolutionary adaptation type inference knowledge extracting apparatus as claimed in claim 5, wherein

the mutational gene manipulating means has a gene replacing means for replacing the arbitrary gene with an arbitrary real number value when executing the gene replacing operation.

7. An evolutionary adaptation type inference knowledge extracting apparatus as claimed in claim 5, wherein

the mutational gene manipulating means has a partial gene string replacing means for replacing a partial gene string subsequent to the arbitrary gene with a partial gene string of another arbitrary individual when executing the gene replacing operation.

8. A fuzzy inference apparatus comprising:

the evolutionary adaptation type inference knowledge extracting apparatus as claimed in claim 1;

an inference input section for inputting numeric data for inference;

a fuzzy inference section for effecting fuzzy inference on the numeric data inputted from the inference input section by means of a fuzzy rule extracted by the evolutionary adaptation type inference knowledge extracting apparatus;

an output section for outputting the fuzzy rule extracted by the evolutionary adaptation type inference knowledge extracting apparatus or a result of the fuzzy inference executed by the fuzzy inference section; and

an inference result correcting section which operates, when correcting the inference result of the numeric data inputted to the fuzzy inference section, to form an input/output data comprised of a combination of the numeric data and corrected data of the inference result, and input the generated data in a feedback manner to an input/output data storing section of the evolutionary adaptation type inference knowledge extracting apparatus.

9. A point of sales data analyzing apparatus comprising:

data format transforming section for transforming a point of sales data into a format of input/output data constituted by a combination of input data comprised of a plurality of numeric values and output data comprised of one discrete value or non-numeric value;

the evolutionary adaptation type inference knowledge extracting apparatus as claimed in claim 1, the apparatus operating, when the input data of the point of sales data transformed by the data format transforming section is inputted, to extract a fuzzy rule representing a tendency of physical distribution included in the point of sales data by updating in an evolutionary adaptation manner a parameter of a fuzzy rule so that an output error with respect to the input data is minimized; and

an inference section for outputting an inference result by executing a fuzzy inference operation with regard to the physical distribution by means of the fuzzy rule extracted by the evolutionary adaptation type inference knowledge extracting apparatus.