关键词: 数据挖掘, 广义粗糙集理论, 广义重要度, 近似约简

Abstract: Considering that the classical rough set theory can only process the discrete data, the degree of general importance of an attribute and attribute subsets was presented. And then a generalization rough set theory was proposed based on the general near neighborhood relation. The theory partitioned the universe into the tolerant modules and formed lower approximation and upper approximation of the set under general near neighborhood relationship which avoided the discretization in Pawlak's rough set. Furthermore, the definition of attribute reduction in generalization rough set and its greedy algorithm were proposed. Finally, results of some examples show the correctness and validity of this method.

Key words: data mining, general rough set theory, degree of general importance, approximation reduction