Research Of Attribute Reduction Algorithm Based On Rough Set

Knowledge representation is the important problem in Artificial Intelligence. In the view of rough set, knowledge is a classification technical, so we can use table to represent knowledge, and divide samples into different types by subset of attributes. While rough set is a technical to deal with uncertain, imprecise and missing information based on divide. Moreover, attribute reduction is the key part in rough set research, is important application of rough set in Intelligent Information Processing, and is important research field in KDD. How to get the effective attribute reduction algorithm, not only is key problem in knowledge reduction algorithm research, but also guard that rough set could be useful in application.In this paper, starting with an introduction to the foundational theory of rough set, and analyzing some important knowledge attribute reduction method, at last focusing on attribute reduction method of Decision Table. In the attribute reduction method of Decision Table, method of Skowron Discernibility matrix is classical, because of which come true easily, so most of modified algorithm based on that.Researches of the paper focus on algorithm of all of attribute reduction set, contents can be classed as follows:1.Research in attribute reduction algorithm in DT, include algorithms of optimal and minimum attribute reduction set, and algorithms of all of minimum attribute reduction set. While analyze their pros and cons, and compare their efficiency.2.In the algorithm of Discernibility function, find some problems influenced efficiency of algorithm, and use Cartesian Product to work out those problems, which effectively increase efficiency of algorithm of Discernibility function in running time and Space Complexity.3.Algorithm of getting relative core based on Discernibility matrix compare with based on attribute importance of Pawlak, by experiment, getting algorithm of getting relative core based on Discernibility matrix is superior.4.Study the Ideal and Boundary algorithm based on Ideal, and demonstrate attribute reduction sets is maximal elements in Ideal sets; While use method of Skowron Discernibility matrix to get the relative core, basing on that, use Boundary algorithm to get attribute reduction sets. In a sense, come up with a new algorithm of attribute reduction.