Research And Application On Variable Precision Rough Set Classification Algorithm

Rough Set theory is a kind of mathematical tools dealing with imprecise,inconsistent and incomplete data. It can obtain knowledge through reduction classification rules. Researches can extract connotative rules from a lot of chaotic data which can be used for the people conveniently. On the application of rough set,we can quickly effectively extract the characteristics of big implied data.But, RS can perform a complete classification requirement,the reduction of attribute set need to be done without information loss,which shows poor antinoise performance. In the real decision-making systems, there exist fuzzy, uncompleted and noisy data.Consequently, the uncertainty of classification and control can’t meet the requirement of RS. In order to overcome this problems, we introduce the Variable Precision Rough Set model. It is one of the most improvement methods in classical rough set model, and it shows the robustness of misclassification and noise in the data concentration. It sets the boundary threshold to relax the stringency regional selection of classical RS by introducing concept of contains degree. Greatly reduced the dimensions of the sample space and the computational complexity of the algorithm.Support Vector Machine is a kind of classification method using widely by academia and industry. It maps the original data to high dimension space by using nonlinear mapping method, and searches the Maximum separation hyperplane in this space. It can minimize the risk structure and has a pretty good generalization performance.This paper deeply studies the Variable Precision Rough Set on the issue of the boundary threshold ? selection,which further improve the fault-tolarent performance of Variable Precision Rough Set model and can be used for medical feature selection.It improves the sample classification accuracy and reduces the time complexity to a certain extent. Finally, applying the SVM to classify the feature set after the screening.In this paper, The main work and innovations in this paper are as follows:1. We put forward the concept of average contains degrees,the condition of average contains degrees as selecting threshold value ? of upper and lower approximation set, according to different types of data sets to generate the optimal variable precision threshold,the domain boundary is large amount of information of condition attribute in positive region.2. For Variable Precision Rough Set reduction based on the approximate constant, this paper further proposes heuristic reduction to solve minimum condition attribute reduction sets. Next, sorting with the condition attribute dependence and selecting in order to save resources and time.3. The above method is applied to extract the medical image features, and using support vector machine to classify the medical image with the screening condition attributes. The experimental results show that the method can generate the optical variable precision threshold according to different types of data sets. Put condition attribute in domain boundary with large amount of information into positive region.