Research On Attribute Reduction Algorithm Based On Decision Tree And Information Entropy

Data mining is a fast, efficient and intelligent analyzing method, which can be used for discovering large amounts of potential information behind the data. Rough set theory is a new math tool to deal with uncertain and imprecision problem. It has been successfully applied to data mining domain. The theory does not need any prior knowledge, it also can simple the expression space of the input information. Under the condition of maintaining the category ability of the knowledge base constantly, attribute reduction deletes redundancy attribute, compacts the constrict rule set in order to help people make the right decision. Searching for efficient algorithm for attribute reduction is the main content of rough set theory. It has important significance in the data mining domain.The constructing algorithm of decision tree based on variable precision rough set and the attribute reduction algorithm based on adaptive particle swarm optimization and information entropy are proposed in the article, respectively.The first algorithm is based on variable precision rough set theory, a new algorithm which uses the new heuristic function construct decision tree is proposed. This algorithm uses variable precision weighted mean roughness as the criterion for selecting categories property. The exact confidence can mark the rules of decision tree in the construction of decision tree, in order to facilitate the comprehension of the decision rule. The corresponding algorithm can be realized on the application of the MATLAB. The validity of the algorithm is proved with an example.The second algorithm is based on the analysis of fuzzy rough set, the reduction algorithm based on adaptive particle swarm optimization algorithm and information entropy is proposed. Fuzzy C mean clustering algorithm based on adaptive particle swarm optimization is used for the analysis of the clustering. The membership matrix which is obtained by the clustering algorithm is used on the attribute reduction. The attribute reduction algorithm based on adaptive particle swarm optimization and information entropy is proposed. Finally, a practical example has been given to verify the practicability of the algorithm.