Two-Phase β-Certain Reducts Generation

摘要

Reducts are applied to represent the knowledge without superfluous attributes in rough set. In this paper, a two-phase β-certain reducts generation is developed to preserve the original classification of each decision class in the table under the majority inclusion relation with a user defined admissible error β. The first phase finds the initial solutions of β-certain reducts. Initial solutions are passed to the second phase and β-certain reducts are found by generating certain reducts of the second pseudo decision when all sub categories based on these certain reducts in the non-β-positive region are totally rejected under the β criterion. No verification is needed when β-certain reducts are found.