An Attribute Reduction Algorithm Based on Maximum External Weight Attribute

In the paper, we firstly analyzed the discrimination ability of condition attribute to individual in universe, then described the discrimination ability of attribute sets to individual using equivalence class partition, finally we defined the internal weight, external weight and the maximum external weight attribute according to the increments of discrimination ability, and established a heuristic search algorithm of condition attribute reduction. The magnitude of the algorithm complexity is polynomial, and the result indicate that the algorithm is not only Pawlak reduction, but also has better convergence.