Attribute Reduction Based on Cross Entropy in Rough Set Theory

摘要

Attribute reduction is very important in rough set-based data analysis because it can be used to simplify the induced decision rules without reducing the classification accuracy. Many types of attribute reductions have been proposed in recent years. However, each of these methods for attribute reduction has their advantages and disadvantages. Cross entropy method can be applied to static and noisy combinatorial optimization problems. In this paper, equivalence partition and relative equivalence partition of domain are processed as probability distributions. Then cross entropy is constructed by two probability distributions in universe. While, relatively dispensable attributes are obtained using cross entropy in this paper. Finally, algorithm of attribute reduction based on cross entropy in rough set theory is presented.