Automatic creation of Neuro-Fuzzy Expert System from online anlytical processing (OLAP) tools

• US 20050149459A1
• Filed: 12/22/2004
• Published: 07/07/2005
• Est. Priority Date: 12/22/2003
• Status: Active Grant

• Current Assignee: Dintecom, Inc.
• Original Assignee: Dintecom, Inc.
• Inventors: Kofman, Gene I., Lobachev, Rouslan V., Lyssenkov, Serguei A.
• Granted Patent: US 7,529,722 B2
• Time in Patent Office: Days
• US Class Current: 706/2
• CPC Class Codes: G06N 5/048 Fuzzy inferencing
• CPC Class Codes: G06N 7/026 Development tools for entering the parameters of a fuzzy system
• CPC Class Codes: G06N 7/046 Implementation by means of a neural network neural networks using fuzzy logic G06N3/043
• CPC Class Codes: Y10S 707/99932 Access augmentation or optimizing

Abstract

A method for automatic generation of a Neuro-Fuzzy Expert System (Fuzzy Logic Expert System implemented as a Neural Network) from data. The method comprising a Data Interface allowing description of location, type, and structure of the Data. The Interface also allows designation of input attributes and output attributes in the Data Structure; automatic Neuro-Fuzzy Expert System generation driven by the Data; Training of the Expert System'"'"'s Neural Network on the Data and the presentation of results which include new knowledge embedded in the parameters and structure of the trained Neuro-Fuzzy Expert System to a user.

Claims

1. A method for automatic generation of a neuro-fuzzy expert system, where a fuzzy logic expert system is implemented as a neural network, the method comprising:

receiving from a data interface selection to describe a location, a type, and a structure of input data, wherein the user interface permits a designation of input attributes and output attributes in a data structure;

generation automatically of a neuro-fuzzy expert system from the input data;

training of a neural network of the neuro-fuzzy expert system with the input data; and

presentation to a user of at least one result representing information embedded in a plurality of parameters and in at least one structure of the neuro-fuzzy expert system which has been trained.

2. The method of claim 1, wherein input data includes:

access to multidimensional data from an OLAP Cube;

designating at least one of a measure for at least one of the input attributes and the output attributes;

designating automatically at least one dimension of the OLAP Cube as input attribute.

3. The method of claim 1, wherein automatic neuro-fuzzy expert system generation includes:

clustering on at least one of input attribute data and output attribute data;

selecting of a result from clustering based on a cluster validity index; and

receiving adjustments from the user into a rule base of the neuro-fuzzy expert system.

4. The method of claim 1, wherein training of an expert system neural network on the input data includes training by:

tuning term sets parameters by at least one of backpropagation and genetic algorithm;

pruning of links in the neural network by a competitive learning algorithm; and

tuning of at least one defuzzification parameter by at least one of the genetic algorithm and a FLC type system.

5. The method of claim 4, wherein the training of an expert system neural network on the data includes tuning of defuzzification parameters by the genetic algorithm utilizing at least one of a distribution of error differential and a combinability relationship during a crossover operation of the genetic algorithm.

6. The method of claim 1, wherein the presentation of the results includes:

displaying logic rules from a rule base inferred during the training of the neural network with multi level sort options; and

displaying at least one of the neural network'"'"'s

nodes,

links, and

weights

utilizing data visualization techniques.

7. A computer program product for automatic generation of a neuro-fuzzy expert system from data, where a fuzzy logic expert system is implemented as a neural network, the computer program product comprising:

a storage medium readable by a processing circuit and storing computer instructions for execution by the processing circuit for performing a method comprising;

receiving from a data interface, one or more selection to describe a location, a type, and a structure of data, wherein the data interface permits a designation of input attributes and output attributes in a data structure;

generation automatically of a neuro-fuzzy expert system from the data;

training of a neural network of the neuro-fuzzy expert system with the data; and

presentation to a user of at least one result representing information embedded in a plurality of parameters and in at least one structure of the neuro-fuzzy expert system which has been trained.

8. The computer program product of claim 7, wherein the data interface includes:

access to multidimensional data from an OLAP cube;

designating at least one of a measure for at least one of the input attributes and the output attributes; and

designating automatically at least one dimension of the OLAP cube as one of the input attributes.

9. The computer program product of claim 7, wherein the data interface includes:

clustering on data corresponding to at least one of the input attributes and the output attributes;

selecting of a result from clustering based on a cluster validity index; and

receiving adjustments from the user into a rule base of the neuro-fuzzy expert system.

10. The computer program product of claim 7, wherein the training of an expert system neural network on the data includes training by:

tuning term sets parameters by at least one of backpropagation and genetic algorithm;

calculating of link weights in the neural network by a competitive learning algorithm;

pruning of links in the neural network by the competitive learning algorithm; and

tuning of a defuzzification parameter by at least one of the genetic algorithm and an FLC type system.

11. The computer program product of claim 10, wherein the training of an expert system neural network on the data includes tuning of defuzzification parameters by the genetic algorithm utilizing at least one of a distribution of error differential and a combinability relationship during a crossover operation of the genetic algorithm.

12. The computer program product of claim 7, wherein the presentation of the results includes:

displaying logic rules from a rule base inferred during the training of the neural network with multi level sort options; and

displaying at least one of the neural network'"'"'s

nodes,

links, and

weights

utilizing data visualization techniques.

13. A system for automatic generation of a neuro-fuzzy expert system from data, where a fuzzy logic expert system is implemented as a neural network, the method comprising:

a data interface for receiving one or more selections to describe a location, a type, and a structure of data, wherein the data interface permits a designation of input attributes and output attributes in a data structure;

an application for automatic generation of a neuro-fuzzy expert system from the data;

a neural network of the neuro-fuzzy expert system trained from the data; and

a user interface for presenting to a user of at least one result representing information embedded in a plurality of parameters and in at least one structure of the neuro-fuzzy expert system which has been trained.

14. The system of claim 13, wherein the data interface includes:

means for accessing multidimensional data from an OLAP cube;

means for designating at least one of a measure for at least one of the input attributes and the output attributes; and

means for designating automatically at least one dimension of the OLAP cube as one of the input attributes.

15. The system of claim 13, wherein automatic neuro-fuzzy expert system generation includes:

means for clustering on data corresponding to at least one of the input attributes and the output attributes;

means for selecting of a result from clustering based on a cluster validity index; and

means for receiving adjustments from the user into a rule base of the neuro-fuzzy expert system.

16. The system of claim 13, wherein the expert system neural network trained on the data includes training by:

tuning term sets parameters by at least one of backpropagation and genetic algorithm;

calculating of link weights in the neural network by a competitive learning algorithm;

pruning of links in the neural network by the competitive learning algorithm; and

tuning of a defuzzification parameter by at least one of the genetic algorithm and an FLC type system.

17. The system of claim 16, wherein the expert system neural network trained by data includes tuning of defuzzification parameters by the genetic algorithm utilizing at least one of a distribution of error differential and a combinability relationship during a crossover operation of the genetic algorithm.

18. The system of claim 13, wherein the user interface for presenting to a user of the results includes:

means for displaying logic rules from a rule base inferred during the training of the neural network with multi level sort options; and

means displaying at least one of the neural network'"'"'s

nodes,

links, and

weights

utilizing data visualization techniques.