A new classification algorithm based on rough set and entropy
Jing Bing Yang, Hao Wang, Xuegang Hu, Zhonghui Hu · 2004
A RSE algorithm for combining rough set theory and entropy heuristics is presented which can induce classification rules, which construction is based on information gain and equivalence relation. The algorithm applies to discrete-valued attributes. So the case of knowledge representation system with some discrete-valued condition attributes and one discrete-valued decision attribute is considered. Firstly, we select a condition attribute based on information gain; secondly, we use rough set theory to establish equivalence classes with respect to the selected condition attribute and decision attribute; finally, classification rules can be extracted from the equivalence classes. Furthermore, we can prove the RSE algorithm valid compared with ID3 algorithm.