A STUDY ON REDUCTION OF ATTRIBUTES BASED ON VARIABLE PRECISION ROUGH SET AND INFORMATION ENTROPY
As a powerful tool for inducing classification knowledge from databases, rough set theory can be used to reduce attributes without the requirement of external information. In previous research, the approximation quality γ is usually used as a criterion in rough set based reduction. But the γ criterion is of limited value when the relationship between attributes is disturbed by noise. Inspired by previous research, this paper proposes an improved criterion for the reduction of attributes based on variable precision rough set and information entropy. Compared with the γ criterion, this criterion could gain more tolerance of inconsistency, randomness and noise. A coefficient of correlation for this criterion indicated by ε is also proposed in order to make the evaluation more reasonable.
Variable precision rough set reduction information entropy
LING SUN JIA-YU CHI ZHONG-FEI LI
College of Lingnan, Sun Yat-sen University, Guangzhou 510275, China School of Management, Sun Yat-sen University, Guangzhou 510275, China
国际会议
2006 International Conference on Machine Learning and Cybernetics(IEEE第五届机器学习与控制论坛)
大连
英文
1412-1416
2006-08-13（万方平台首次上网日期，不代表论文的发表时间）