Research On The Attribute Reduction In Inconsistent And Incomplete Decision System

Z.Pawlak, a polish mathematician professor, put forward a new mathematical tool——Rough Set Theory, which employed in processing incomplete, uncertain and imprecise information in 1982. Nowadays, the theory has been successfully applied to the fields of artificial intelligence, pattern recognition, intelligent information processing etc. As a critical process in the method of knowledge acquisition, attribute reduction is one of the core issues in Rough Set theory’s research. Attribute reduction is to maintain the data in the original message unchanged, and removes some of the irrelevancy and redundant attributes.In reality, in the view of anthropogenic causes or external environment etc, the finally obtained decision systems are often inconsistent. This paper mainly studies the problem of attribute reduction in inconsistent decision system, focusing on the solutions to assignment reduction in inconsistent incomplete decision system and attribute reduction in test-cost-independent decision system.The main research work of this paper focus on the following aspects:(1) Due to the fact that assignment reduction solution of the existing inconsistent incomplete decision system exists the problems of high space complexity, which is not suitable for large-scale databases and other issues. An assignment reduction’s solving method based on the conflict object set in inconsistent incomplete decision system is designed in this paper. Via presenting the concept of conflict object set, the methods for computing core attribute set and measuring attribute significance are given. Based on these conditions, an assignment reduction is proposed to solve algorithm in inconsistent incomplete decision system. The theoretical analysis and example analysis results show that the proposed method in this paper significantly reduces the space complexity of assignment reduction, which is more suitable for large-scale databases.(2) For the case of test-cost-independent decision system exists conflict object, in order to decrease test cost concerning attribute reduction. In this paper, an attribute reduction method in test-cost-independent decision system based on conflict region is proposed. First of all, simplify decision system which exists conflict object, and then, according to the degree of conflict intensity calculate the core attribute and attribution importance on the simplified decision system. Based on these conditions, a heuristic function to solve the low test cost’s attribute reduction is designed. Experimental results on UCI(University of California Irvine) datasets show that the proposed approach in this paper remarkably improves the efficiency of attribute reduction in test-cost-independent decision system.