A new knowledge reduction method based on difference-similitude set theory

A new knowledge reduction method based on difference-similitude set theory (DSST) is proposed. It describes each originating rule with the corresponding difference set DS and similitude set SS. Elements in DS are divided into 3 classes: base set, no-base set and cover-base set. Attributes only in the cover-base set can be rejected from IS immediately while computation of DS is accomplished and the remained reduction is only processed in no-base set. The reducing process is based on the implicated knowledge in DS and SS: which attribute can be rejected or not, is determined by DS (every element in DS should not be empty) and which attribute to reject is more reasonable (optimal) is determined by SS (the corresponding overlapness should be the minimum one). Being a heuristic search algorithm in reducing knowledge, it takes not only the difference but also the similitude in IS into account. Because the reduction is only processed in no-base set and in the restriction of optimal principle, the complexity of the computation is far decreased and the efficiency is far increased.