Complexity reduction in fuzzy modeling

The interest in data driven approaches to the acquisition of fuzzy systems is increasing. Most of the approaches in the literature emphasize the global quantitative accuracy and not the transparency and interpretability of the resulting model. This paper discusses methods based on similarity analysis that, without performing additional knowledge or data acquisition, allow for the generation of fuzzy models of varying complexity. While models for simulation emphasize numerical accuracy, models for understanding the system and for operator interface are required to be transparent and interpretable. An application of the presented fuzzy modeling techniques to an air-conditioning system is described.