Rough set based feature selection for improved differentiation of traditional Chinese medical data

Medical data often contains a large number of irrelevant and redundant features and a relatively small number of cases, which dramatically impact quality of diseases diagnosis. Hence, in quest for higher differentiation quality, feature selection is expected to improve differentiation performance. In this paper, we describe a heuristic approach based on Rough Sets theory and information theory, for generation of a reduct approximation of a medical dataset. The algorithm consists of two phases: initializing starting point phase and heuristic search phase. The experimental results on the medical datasets of UCI machine learning repository and traditional Chinese medicine datasets show that the proposed algorithm can efficiently select critical features and improve the performance of differentiation.