[[alternative]]An Approximation Reasoning Approach for Generating Fuzzy Decision Rules

[[abstract]]Most fuzzy classification systems proposed before applied a crisp-cut approach on the fuzzy degrees of the fuzzy attributes and conclusions to generate decision rules. Although, by the crisp-cut approach, decision rules with conjunction-disjunction form can be derived from training-samples, the membership functions of the conclusions cannot be generated. In this paper, a learning method named Fuzzy Approximation Reasoning Method is proposed. Two requirements can be satisfied by the method：(1)deriving fuzzy decision rules with conjunction-disjunction form from training-samples, and (2)generating the membership functions for the conclusions. In Fuzzy Approximation Reasoning Method, the dependency-degree function is designed for estimating the relationship between a conclusion and the fuzzy attributes. For the fuzzy attributes related to the conclusion, their membership functions will be combined to construct the membership function of the conclusion such that the associated fuzzy decision rule is derived. Moreover, the Fuzzy Approximation Reasoning Method also can be used to mine fuzzy association rules. In this paper, the Approximation Inducing Method is proposed to demonstrate how to mine fuzzy association rules by applying the Fuzzy Approximation Reasoning Method.