2nd Workshop on Computing and Communications from Biological Systems: Theory and Applications

Research Article

Evolutionary Multiobjective Fuzzy System Design

  • @INPROCEEDINGS{10.4108/ICST.BIONETICS2008.4718,
        author={Hisao Ishibuchi and Yusuke Nojima},
        title={Evolutionary Multiobjective Fuzzy System Design},
        proceedings={2nd Workshop on Computing and Communications from Biological Systems: Theory and Applications},
        publisher={ACM},
        proceedings_a={CCBS},
        year={2010},
        month={5},
        keywords={Fuzzy rules fuzzy systems genetic algorithms evolutionary multiobjective optimization multiobjective design.},
        doi={10.4108/ICST.BIONETICS2008.4718}
    }
    
  • Hisao Ishibuchi
    Yusuke Nojima
    Year: 2010
    Evolutionary Multiobjective Fuzzy System Design
    CCBS
    ICST
    DOI: 10.4108/ICST.BIONETICS2008.4718
Hisao Ishibuchi1,*, Yusuke Nojima2,*
  • 1: Osaka Prefecture University 1-1 Gakuen-cho, Naka-ku Sakai, Osaka 599-8531, Japan +81-72-254-9350
  • 2: Osaka Prefecture University 1-1 Gakuen-cho, Naka-ku Sakai, Osaka 599-8531, Japan +81-72-254-9198
*Contact email: hisaoi@cs.osakafu-u.ac.jp, nojima@cs.osakafu-u.ac.jp

Abstract

This paper briefly reviews genetic algorithm-based approaches to the design of fuzzy systems. In the 1990s, genetic algorithms were mainly used for the accuracy maximization of fuzzy systems. Various aspects of fuzzy systems were optimized by genetic algorithms such as the fuzzy partition of each input variable, the number of fuzzy rules, and the consequent part of each fuzzy rule. The accuracy maximization of fuzzy systems for training data, however, tends to increase their complexity. That is, the accuracy maximization often degrades the interpretability of fuzzy systems through the increase in their complexity. Some studies in the late 1990s tried to find a good tradeoff (i.e., compromise) between the accuracy and the complexity of fuzzy systems. The latest trend in the design of fuzzy systems is their evolutionary multiobjective design. A number of non-dominated fuzzy systems with different accuracy-complexity tradeoffs can be obtained by a single run of multiobjective approaches. In this paper, we briefly review the above-mentioned main stream of research on fuzzy system design.