Attribute Reduction And Empirical Analysis Based On The Granularity Computing

Attribute reduction is an important research field in the granular computing of rough set theory. Three kinds of attribute reduction methods are discussed in this paper. The data in the stock market is discretized by rough set theory and reduced by attribute reduction methods, and making some analysis for the stock market, and results of cases show that the methods are feasible, and providing a new way to study the stock marketThe concrete research work is as follows:1. The basic knowledge of rough set and fractal market are briefly introduced in the second chapter, and introducing attribute reduction algorithm based on the roughness and giving examples and a detailed statistical analysis and fractal research, attribute reduction algorithm which is combined with fractal market is obtained, injecting new vitality for the stock market.2. Several currently used methods of attribute reduction in rough set are briefly introduced in the third chapter, and introducing attribute reduction based on the information entropy, then reducing the data in the stock market by this method. The result of case shows that each attribute is important.3. Several methods of data discretization are briefly introduced in Chapter IV, and then focus on introducing dynamic hierarchical cluster, then re-discreting the data of the stock market by this method. The consequences of cases show that the same reduction algorithm facing to different discretization methods will have different reduction results and that the same discretization method facing to different reduction algorithm will have different reduction results, too.A consequence can be obtained from this paper:the rough set is an effective tool. The redundant attributes in data in the stock market are deleted by attribute reduction, which provides a new approach for analysis and decision-making for data in the stock market.