Heuristic Algorithm for Attribute Reduction Based on Classification Ability by Condition Attributes
Yasuo Kudo, Tetsuya Murai · Journal of Advanced Computational Intelligence and Intelligent Informatics · 2011
The heuristic algorithm we propose to compute a relative reduct candidate is based on evaluating classification ability of condition attributes. Considering the discernibility and equivalence of objects for condition attributes in relative reducts, we introduce evaluation criteria for condition attributes and relative reducts. The computational complexity of the proposed algorithm isO(|U|2|C|2). Experimental results indicate that our algorithm often generates a relative reduct producing a good evaluation result.