Study On Cluster Analysis And Rule Mining Based On Granular Theory

With the development of information technology, data mining, as a new kind of information processing technology, has becoming to be mature gradually. Data mining is the advanced process of discovering some latent ,interesting and valuable knowledge from large amount of data o Data clustering analysis (DCA) and association rule mining are both the effective methods of information processing technology in the fields of data mining. Data clustering analysis is an important research field of data mining. And it is an important method for data partition or group dealing.Data clustering analysis can also be seen as a data preprocessing technique, the unknown but valuable information can be discovered from massive data by means of DCA, which provide powerful support for the efficient information analysis and processing. Association rule mining can reveal the unknown rules hidden in large databases, which brings advice for human activities. Granular computing is a simulation of human intelligence, human can think, analyze and solve the problem in different sizes, so granular computing takes advantage of this particular human feature for dealing with complex, unstructured, incomplete, uncertain information. Hierarchy is an important concept of granular thinking mode. As a new method of dealing of knowledge, it already has been widely applied to the field in automatic control and management decision-making.This thesis firstly presented the basic knowledge of data mining, and discussed the theory and characteristics of data cluster analysis and association rule mining. Then we applied the theory of hierarchical problem-anglicizing and solving in the fields of granular computing to the data clustering algorithm to generate a more efficient two-stage hierarchical clustering algorithm based on granular theory. The theoretical analysis and the simulation results can approved that the proposed algorithm improved a certain extent in time and space performance than the traditional hierarchical clustering algorithm in dealing with high-dimensional and large-scale data samples. In the end, according to the above research on classic association rule mining algorithm in details, through summarizing its limits and studying the advantages of information granules in dealing with hierarchical data and frequent item set, we proposed a rule mining algorithm based on granular computing. The result of an experiment proved the algorithm of rule mining based on granular computing is feasible and effective.