ABSTRACT
Fuzzy decision trees exploit popularity of decision tree algorithms for practical knowledge acquisition and representative power of fuzzy representation. They are extensions of symbolic trees, with tree-building routines modified to utilize fuzzy instead of strict domains, and with new inferences combining fuzzy interpolation and defuzzification with inductive methodology. This chapter describes a method for optimizing the fuzzy component of knowledge represented in the fuzzy tree. That is, it optimizes the knowledge by adjusting the fuzzy sets for both linguistic terms and for decisions, and by selecting an optimal norm for combining conjunctions of fuzzy restrictions. A genetic algorithm is used to perform the highly constrained optimization. Fuzzy sets are one of the most popular methods designed to overcome such limitations of symbolic systems. They provide bases for fuzzy representation. Decision trees are made of two major components: a procedure to build the symbolic tree and an inference for decision making.
