tailieunhanh - Fuzzy distance based attribute reduction in decision tables

In this paper, we propose a fuzzy distance based attribute reduction method on the decision table with numerical attribute value domain. Experiments on data sets show that the proposed method is more efficient than the ones based on Shannon’s entropy on the executed time and the classification accuracy of reduct. | Các công trình nghiên cứu phát triển và ứng dụng CNTT-TT Tập V-2 Số 16 36 tháng 12 2016 Fuzzy Distance Based Attribute Reduction in Decision Tables Cao Chinh Nghia Vu Duc Thi Nguyen Long Giang Tan Hanh Abstract In recent years fuzzy rough set based attribute reduction has attracted the interest of many researchers. The attribute reduction methods can perform directly on the decision tables with numerical attribute value domain. In this paper we propose a fuzzy distance based attribute reduction method on the decision table with numerical attribute value domain. Experiments on data sets show that the proposed method is more efficient than the ones based on Shannon s entropy on the executed time and the classification accuracy of reduct. Keywords Fuzzy rough set fuzzy decision table fuzzy equivalence relation fuzzy distance attribute reduction reduct. I. INTRODUCTION Attribute reduction is an important issue in data preprocessing steps which aims at eliminating redundant attributes to enhance the effectiveness of data mining techniques. Rough set theory 12 is an effective approach to solve feature selection problems with discrete attribute value domain. Traditional rough set based attribute reduction techniques have many limitations when performing on tables with numerical attribute value domain. Data needs to be discretized before performing attribute reduction techniques. The major limitation of rough set theory based attribute reduction is losing information in the discrete processing which will affect the quality of data classification. To solve the problem of attribute reduction directly on decision table with numerical data fuzzy rough set based approach has recently been developed 3-6 10 16 17 . Dubois D. and Prade H. proposed fuzzy rough set theory 3 4 which is a combination of rough set theory 12 and fuzzy set theory 18 in order to approximate fuzzy sets based on fuzzy equivalence relation. In rough set theory two objects are called equivalent on R attribute