An algorithm for sub-optimal attribute reduction in decision table based on neighborhood rough set model

摘要

In this paper, some concepts of upper approximation and lower approximation and so on are defined concisely and strictly on neighborhood rough set model. According to the fruit fly optimization algorithm's idea, an new algorithm(NBH SFR) to get a sub-optimal attribute reduction on neighborhood decision table is proposed. The validity and feasibility of the algorithm are demonstrated by the results of experiments on four UCI Machine Learning database. A detailed analysis of δ operator to influence on the results is given. And the δ operator formula to obtain a sub-optimal reduction is proposed. Moreover, the experiments also show that it is impossible to solve multi-dimensional big dataset based on kernel-based heuristic algorithm ideas.