Induction of fuzzy decision tree based on fuzzy similarity relationship

A method based on fuzzy similarity relationship is proposed for learning a fuzzy decision tree from a real-valued decision system. First, fuzzy similarity relationships are utilized to fuzzy the real-valued attributes of the decision system into fuzzy-valued attributes by computing the closures of the similarity matrices. Second, a measure of importance of condition attributes is introduced using the loose lower approximation, which is based on similarity relationship. Finally, using the measure as the criteria to select the expanding attribute, an algorithm for a fuzzy decision tree is proposed, and an illustrative example is demonstrated.