Information-preserving rule induction by using generalized fuzzy-rough technique

In this paper, we build a rule-based classifier by using generalized FRS with variable precision after proposing a new, concept named 'consistence degree' which is used as the critical value to keep the information invariant in the processing of rule induction. First, we improve the existing FRS by incorporating one controlled threshold into knowledge representation of fuzzy rough sets so that fuzzy rough sets become a robust model. Second, we describe some concepts of attribute-value reduction. The key idea of attribute-value reduction is to keep the consistence degree, i.e. fuzzy lower approximation value of certain decision class invariant before and after reduction. Third, a set of rules which covers all the objects in the original dataset can be obtained after the description of rule representation system in fuzzy decision table. Finally, the experimental results show that (he proposed rule-based classifier is feasible, and effective on noisy data. The main contribution of this paper is that the rule induction method is well combined with knowledge representation of fuzzy rough sets by using fuzzy lower approximation value.