Automatic generation of first order Takagi–Sugeno systems using fuzzy c-elliptotypes clustering

Abstract

The lack of skilled experts being able to provide objective knowledge about complex systems motivates the development of automatic tools for the extraction of fuzzy if-then rules from available data sets. Fuzzy clustering is a well-established and widely used method for data analysis. In this paper, fuzzy c-elliptotypes clustering is used to extract locally linear models, which are translated into first order Takagi–Sugeno rules. The main goal is the extraction of good fuzzy models containing human interpretable information (readable transparent rule bases) from high-dimensional data sets. Existing methods as well as new approaches are used to automatically set the cluster parameters, to reduce the dimensions of the clusters and to generate approximate membership functions. The experiments show that the proposed methods are well suited for complex real-world problems.