No label defined (Q4241983)

The paper forms a comprehensive study of the fuzzy rule-based models and establishes a thorough taxonomy of such models. The authors discuss the underlying mechanisms of approximate reasoning used throughout this form of fuzzy modeling. The underlying architecture of the rule-based models dwells on ``if-then'' statements of the form \[ \text{if }U\text{ is }B_i\text{ then }V\text{ is }D_i,\quad i=1,2,\dots,c, \] where \(B_i\) and \(D_i\) are fuzzy sets (relations) defined in the input \(({\mathbf U})\) and output \(({\mathbf V})\) space, respectively. The basic schemes of inference carried out in fuzzy models are presented (both constructive and destructive ways of inference are discussed and contrasted). Various models of decoding (defuzzification) of fuzzy sets of output are discussed. The role of aggregation operations in fuzzy modeling is also presented in detail. The ensuing analysis deals with some fundamental properties as well as elaborates on an impact of the aggregation mechanisms on the performance of the models. For the entire collection see [Zbl 0911.00016].