Special Issue on Genetic Fuzzy Systems and the Interpretability-Accuracy Trade-off

This special issue encompasses four papers devoted to the recent developments in the field of ‘‘Genetic fuzzy systems and the trade-off between interpretability and accuracy’’. The issue was originated from several contributions presented at the First International Workshop on Genetic Fuzzy Systems (GFS2005) that was held in Granada, Spain, March 17–19, 2005. Six conference papers were selected and the authors were asked to develop extended versions which were submitted to the special issue. Each of them was revised by at least three referees and finally four of them were accepted according to the reviewers’ evaluations. System modelling with fuzzy rule-based systems comes with two contradictory requirements in the obtained model: the interpretability, capability the behaviour of the real system in an understandable way, and the accuracy, capability to faithfully represent the real system. Obtaining high degrees of interpretability and accuracy is a contradictory purpose and, in practice, one of the two properties prevails over the other. While linguistic fuzzy modelling (mainly developed by linguistic or Mamdani-type fuzzy systems) is focused on the interpretability, precise fuzzy modelling (mainly developed by Takagi–Sugeno–Kang fuzzy systems) is focused on the accuracy. The relatively easy design of fuzzy systems, their attractive advantages, and their emergent proliferation have made fuzzy modelling suffer a deviation from the seminal purpose directed towards exploiting the descriptive power of the concept of a linguistic variable. Instead, in the last few years, the prevailing research in fuzzy modelling has focused on increasing the accuracy as much as possible, paying little attention to the interpretability of the final model. Nevertheless, a new tendency in the fuzzy modelling scientific community that looks for a good balance between interpretability and accuracy is increasing in importance. In order to successfully get this interpretability–accuracy trade-off, the use of genetic algorithms to search optimal fuzzy models is especially helpful. Indeed, they have a powerful search capability that allows them to deal with difficult objective functions and to perform multiobjective optimization. Moreover, they can deal with flexible representation structures with mixed coding schemes and constrains. These features are explored by the