The Research Of Data Analysis Methods And Applications Based On Rough Set

Rough set theory, proposed by Z. Pawlak during the early 1980s, is a new approach to handling imperfect data (uncertainty and vagueness). It further complements other theories that deal with data uncertainty, such as probability theory, Dempster-Shafer theory of evidence theory, fuzzy set theory, etc. After nearly 30 years of development, rough set model has been widely used in many fields. The remarkable advantage of rough set theory is that it can discover the dependency hidden in data, without any preliminary or additional information about data. A lot of interesting applications of the rough set can be found. Nowadays, the theory has become to be one of the most important mathematical tools in the area of intelligent information processing. A large number of published papers or literatures show that: data reduction, evaluation of the significance of data, and decision rules generation from data are the kernel research contents of rough set theory and its applications. It sometime helps improve the efficiency of data analysis that using data reduction to eliminate irrelevant or redundant data. With the research development of rough set theory and its applications in recent years, the fusion model combining rough set approach with other methods has gradually attracted researchers'attention.In this article, the background of rough set theory is firstly introduced. And the basic concepts of rough set theory are also briefly presented in chapter 2. Three topics are discussed in the later, such like discretization algorithm of continuous attributes, weights design in multi-attribute decision making problems, and support vector regression model. Drawing on the experience of previous studies, rough set theory is applied to the above problems. The main works of this article are as follow:(1) Analyzing the process of SOM algorithm, then Using feedback information from decision table consistency, a new data discretization algorithm based on dynamical SOM cluster algorithm is proposed. This algorithm not only can be applied in consistent decision tables, but also inconsistent decision tables. Data experiments show that, compared with some other data discretization algorithms, the new discretization algorithm's performance is better. (2) Introduce the concept of rough set fuzziness proposed by Wang Guo-Yin et al. to apply for measuring the uncertainty of rough set. A new design proposed in chapter 4:integrate attribute reduction algorithm and the weights measuring method based on the fuzziness of rough set with TOPSIS model, and extended to multi-attribute group decision making approach. Some numerical examples for simulation and experiment are given to demonstrate the validity, convenience and applied values of the proposed design.(3) Integrate attribute reduction algorithm with support vector regression machine to bring forward a new model RSSVR, which is applied in forecast of China's power supply. Case study shows that the RSSVR model has a better forecast performance.