The Research Of Attribute Reduction Algorithm Based On Rough Set Theory

Since the 1980s, along with the electronic commerce, as well as the rapid popularization of automation in government and enterprise affairs, a mass of data is accumulated in information systems; "data rich and information poor" have become a great challenge of the digital community. In the vast and diverse data, how to mine potential and valuable information is the main research topics of data mining. Rough Set Theory, which is proposed by Polish mathematician Z. Pawlak in the 1982, is a new mathematical tool studied for fuzziness and uncertainty. Under the premise of maintain invariability of the classified ability, without pre-setting the number description of certain characteristics or attributes but directly from a given set of a description of the problem, educed the classification rules of concept through knowledge reduction is the important feature of rough set. An important means of knowledge discovery and extracting rules is knowledge reduction. The current knowledge reduction algorithm is based mainly on the Rough Set Theory, in the face of a huge amount of data, effective and rapid algorithm is crucial.Firstly, the research actualities at home and abroad of rough set theory are summed up in the paper, whose relevant concepts, working processes and pivotal technologies are discussed. Moreover two rapid attribute reduction algorithms are proposed based on equivalence relation:1) Firstly, combining with granular computing and rough set theory, the granular decompounding method along with the computing ways of positive and negative granular region in decision table system are given, further more the attribute reduction algorithm based on granular expression is propose, giving the attribute importance as heuristic information. Analysis shows that the time complexity of this algorithm is smaller, and the experimental result proves that the proposed algorithm not only can reduce the decision table efficient but rapid.2) In order to avoid the genetic algorithm geting into partial optimum, the second algorithm using immune thinking joined into genetic algorithm, and the core is introduced the initial antibody colony of immune genetic algorithm to improve capability. According to dependence of decision attribute vs. condition attribute, combining with the consistency of antibody, the attribute reduction algorithm is proposed based on immune genetic algorithm. The experiment result shows that it can reduce decision table effectively and converge rapidly in the global optimum.