Select Representative Samples for Regularized Multiple-Criteria Linear Programming Classification

Peng Zhang, Yingjie Tian, Xingsen Li, Zhiwang Zhang, Yong Shi

Abstract: Regularized multiple-criteria linear programming (RMCLP) model is a new powerful method for classification in data mining. Taking account of every training instance, RMCLP is sensitive to the outliers. In this paper, we propose a sample selection method to seek the representative points for RMCLP model, just as finding the support vectors to support vector machine (SVM). This sample selection method also can exclude the outliers in training set and reduce the quantity of training samples, which can significantly save costs in business world because labeling training samples is usually expensive and sometimes impossible. Experimental results show our method not only reduces the quality of training instances, but also improves the performance of RMCLP.

External IDs:dblp:conf/iccS/ZhangTLZS08