A Data Preprocessing Algorithm Based on Rough Set for SVM Classifier

Support vector machine (SVM) is now widely applied in various areas for its excellent performances. For a data set, usually we use normalization method to deal with the features. However, in many cases, the value of each feature is different. Thus, SVM can't work very well. In this paper, we propose a preprocessing algorithm based on rough set (RS) theory to give different weights on each feature, which can well reflect the value of each feature. The experimental results on real data show that the proposed approach can achieve a fairly improvement of classification accuracy.