Evolving Takagi-Sugeno Fuzzy Models for Data Mining

Authors

Abstract

We are studying and experiencing approaches for adaptive on-line learning of fuzzy rules and their application for prediction problems in the context of data mining. Takagi-Sugeno (TS) fuzzy models are used for knowledge representation and the mechanism for on-line learning is based on algorithms that recursively update the model structure and parameters by combining supervised and unsupervised (hybrid) learning. The structure and parameters of the model continually evolve by adding new rules and by modifying existing rules and parameters during the operation of the system. The work is based on developments from the original contributions of Stephen Chiu, Plamen Angelov and Nikola Kasabov.

Keywords

Takagi-Sugeno fuzzy models, fuzzy clustering, rule-base adaptation, on-line learning

Subject

On-line Learning Algorithms

Conference

SoftComplex'03, June 2003