Rule Extraction Using Soft-Computing Techniques with Fuzzy Linguistic Representation

Abstract

Decisions for real-world problems are not always made precisely since the input data are themselves imprecise. This study presents a rough-fuzzy hybridization method to generate fuzzy if-then rules automatically from a diagnosis dataset with quantitative data values, based on fuzzy set and rough set theory. The proposed method consists of four stages: preprocessing inputs with fuzzy linguistic representation; rough set theory in finding notable reducts; candidate fuzzy if-then rules generation by data summarization, and truth evaluation the effectiveness of fuzzy if-then rules. The main contributions of the proposed method are the capability of fuzzy linguistic representation of the if-then rules, finding concise fuzzy if-then rules from diagnosis dataset, and tolerance of imprecise data.