Similarity matrix-based algorithm for reduction of attribute in incomplete information systems

Aiming at the matter of the heavy workload of discernibility matrix and the inefficient of the reduction strategy based on discernibility matrix, this paper proposes a new attributes reduction arithmetic-based similarity matrix. In this paper, We firstly constructs the correlation between similarity relation and matrix, puts forward a new index of attribute significance, and used the new index as heuristic information to design a new attribute reduction algorithm. The algorithm proposed in this paper can avoid repeatedly calculation important value of unimportant attributes and decrease the temporal complexity.