RESEARCH ON ATTRIBUTE REDUCTION ALGORITHM IN COVERING GRANULAR COMPUTING MODEL

The granular computing model is a very important model in data mining. Seven kinds of models have been proposed for the covering granular computing model. Among these models, the third kind of models was considered to be more reasonable model. However, there are not many solutions to attribute reduction based on the third kind of covering models. Aiming at the third kind of covering granular computing models, this paper discusses the concepts of covering related families and simplified covering related families, and gives a simplified calculation method for covering related families, and then calculates all attribute reductions from the simplified covering related families. Then, a heuristic method of attribute reduction is discussed. The feasibility of the proposed method is verified by an example, which provides a new method for attribute reduction in data mining.