A Novel Attribute Reduction Algorithm of Decomposition Based on Rough Sets

Attribute reduction is a key task for the research of rough sets. However, when dealing with large-scale data, many existing proposals based on rough set theory get worse performance. In this paper, we propose a novel attribute reduction algorithm of decomposition based on rough sets. The idea of decomposition is to break down a complex table into a super-table and several sub-tables that are simpler, more manageable and solvable by using existing induction methods, then joining them together in order to solve the original table. Compared with the traditional methods, experiments with some standard datasets from UCI database are done and experimental results illustrate that the algorithm of this paper improve computational efficiency.