Attribute Reduction Algorithm Based on Relative Granularity in Decision Tables

A relative attribute significance of decision tables,based on the theory of knowledge granularity,was defined by introducing the concept of relative granularity,and the relative granularity’s monotonous increasing property with the increase of knowledge granularity was proved,then a heuristic reduction algorithm based on relative granularity was proposed.The algorithm which eliminates the limitation of reduction algorithms based on positive region in dealing with inconsistent decision table,by analyzing essential theory and application examples,is proved as effective,and it’s time complexity is relatively low.