Research On Rough Set Theory In Dicision System

Data Mining which research how to discovery valuable knowledge and rules from large scale database, is one of the active sub area of Artificial Intelligence with a broad application in many business and scientific area. Rough set theory, a mathematical tool for mining rules from uncertain information system, possesses strong abilities in dealing with uncertain information system. Based on rough set theory, research in this paper mainly focuses on mining rules from decision system.Based on the research of the basic concepts of rough set, we deeply analysis and discuss the approximation set algorithm, equivalence class algorithm, core algorithm, attribute reduction algorithm, rules extract algorithm etc. And for every algorithm, we discuss the implement method in detail.Attributes reduction is an important concept and also, attributes reduction algorithm plays a very important role in rough set theory. Because of high exponential time complexity of attributes reductionalgorithm such as traditional reduction algorithm and disernibility matrix algorithm, this paper propose a new attributes reduction algorithm, which starting from core attributes set and taking power set of attributes set as a computational tool, removes the useless attribute one by one, according to the cardinality of elements in the power set. The algorithm analysis shows the new attribute reduction algorithm lower the time complexity to polynomial level. Based the new reduction algorithm, we propose a heuristic reduction algorithm, which lower the time complexity further.We also present a new equivalence class algorithm, which scan decision table only once and generate all the equivalence class. Analysis shows this algorithm is efficient in large scale decision system.