Rough Entropy-based Feature Selection and Its Application

摘要

Feature selection from granular computing is a basic issue in knowledge representation and data mining. In this paper, the boundedness of the recent feature selection algorithms were first discussed, then a new rough entropy to measure the roughness of knowledge and its properties were proposed in decision information system, and the conclusion that rough entropy decreased monotonously as the information granularities became finer was obtained. Thus, the new significance of feature and the heuristic algorithm of feature selection were proposed. Finally, through analyzing the given example, it was shown that the proposed heuristic information was better and more efficient than previous schemes, and numerical simulation experiments from gene expression data sets showed that the proposed approach was practical and effective.