Maximum condition entropy based attribute reduction in variable precision rough set model

Variable precision rough set model, as a probabilistic extension of original rough set model, is a very useful approach to inducing probabilistic rules from datasets. In this paper, some anomalies in present definition of attribute reduction based on variable precision rough set model are discussed. Maximum condition entropy is introduced to analyze the mergers of condition classes in the process of reduction and construct a monotonic measure for attribute reduction. Then, based on core attributes under maximum condition entropy, a new heuristic algorithm is proposed to compute reduct, which eradicates all anomalies in variable precision rough set model. Finally, an example is used to show the validity of proposed algorithm.