Evolution of Fuzzy Rule Based Classifiers

The paper presents an evolutionary approach for generating fuzzy rule based classifier. First, a classification problem is divided into several two-class problems following a fuzzy unordered class binarization scheme; next, a fuzzy rule is evolved (not only the condition but the fuzzy sets are evolved (tuned) too) for each two-class problem using a Michigan iterative learning approach; finally, the evolved fuzzy rules are integrated using the fuzzy round robin class binarization scheme. In particular, heaps encoding scheme is used for evolving the fuzzy rules along with a set of special genetic operators (variable length crossover, gene addition and gene deletion). Experiments are conducted with different public available data sets.