A SVM Classifier Research Based on RS Reducts

Support vector machine (SVM) is a promising method of machine learning based on the structural risk minimization principle, which is characteristic of good generalization performance; Rough set (RS) is an effective tool to decrease data dimension in dealing with vagueness and uncertainty information. A SVM classifier based on RS reducts is researched in order to enhance the predicting performance in this paper. We firstly get all the attribute reducts; Then, every element is chosen from the set of all reducts and relevant SVM classifiers are constructed respectively; Finally an enhanced SVM classifier based on RS reducts is obtained from them, which make great use of the advantages of RS in eliminating redundant information and take full advantage of SVM to train and test the data. Experiment results explain the validity and feasibility of our proposed algorithm.