Entropy-based fuzzy rough classification approach for extracting classification rules

摘要

Recently, the combination of Fuzzy Set Theory and Rough Set Theory has become a popular data mining technique for classification problems because of their strength of handling vague and imprecise data. From the previous literature, Rough Set Theory can only operate effectively with datasets containing discrete values. As most datasets contain real-valued features, it is necessary to perform a discretization step beforehand, which is typically implemented by standard fuzzification techniques. In this paper, a new fuzzification technique called Modified Minimization Entropy Principle Algorithm (MMEPA) is proposed to construct membership functions of fuzzy sets of linguistic variables. Using the dataset fuzzified by this technique to perform the rule extraction algorithm Variable Precision Rough Set Model (VP-model), the extracted classification rules by this model can obtain a higher classification accuracy rate than that of some existing methods.