Rough Sets Based Research On Method Of Discretization And Reduction Algorithm

Rough set theory proposed firstly by Polish mathematician Z.Pawlak in 1982 is a valid mathematical tool to handle uncertainty after Probability theory and fuzzy set theory. Rough set theory is based on the indiscernibility relation that describes indistinguishable objects and can be approached by two accurate sets, the lower and upper approximation. Not needing other information this theory can analyze and process the imprecise, uncertain and incomplete data problems. In recent years,rough sets theory has been successfully implemented in Data Mining, Artificial Intelligence, Pattern Recognition, Rough Control, etc.Continuous data discretization is one of important problems in rough set theory. In many real-life applications often include continuous data and Rough set theory proposed by Z.Pawlak based on indiscernibility only can deal with discrete attribute value, so how to discretize continuous data becomes the bottleneck of using rough set theory. The key problem of discretization is to seek appropriate and consistent cut points in the space of condition attributes and is a space partition and code optimization problem. In this dissertation, it introduces many kinds of discretization methods firstly and then studies a heuristic genetic algorithm, based on that I improved this method by judging the significance of cut points to help the mutation in gene evolution and adding a modify progress which ensure the feasibility of searching result. After using the Iris dataset for experiment, we proved that this new algorithm is effective than the old one.Attribute reduction is one of main topics in rough set theory. The reduction of attributes can highly enhance the efficiency of data processing by removing redundant conditional attributes and keep the best decision, which has significance in real-world. Getting the best attribute reduction and rule reduction are NP-Hard Problems which is proved by Wong SKM and Ziarko W. So how to study more effective algorithms to get the better attribute reduction and to decrease the time complexity becomes an important point in rough set theory. Exiting literature about the attribute reduction algorithms are mostly based on discernibility matrix method and attribute significance. After discussing these familiar approaches, this paper proposed a new attribute reduction algorithm in rough set combined fuzzy relation theory by importing fuzzy relation and compound matrices computing. According to the decision and domain knowledge, the user can set threshold value to get the satisfying result. By using family cars dataset we proved that this algorithm is useful and effective.In the Chapter Six the dissertation discussed about how to combine fuzzy set and rough set theory together, hybrid system and research methodology about rough set which is the further research plan and work in future about this field.