Density-Based Penalty Parameter Optimization on C-SVM

• DOI: 10.1155/2014/851814
• Publication Date: 2014-07-07T17:07:35Z
• Authors: Yun Liu; Jie Lian; Michael R. Bartol...; Qing-An Zeng
• Keywords: Interface (matter); Data set; Ranking SVM

The support vector machine (SVM) is one of the most widely used approaches for data classification and regression. SVM achieves largest distance between positive negative vectors, which neglects remote instances away from interface. In order to avoid a position change interface as result an error system outlier, C-SVM was implemented decrease influences system's outliers. Traditional holds uniform parameter C both instances; however, according different number proportions distribution, should be set with weights penalty terms. Therefore, in this paper, we propose density-based optimization C-SVM. experiential results indicated that our proposed algorithm has outstanding performance respect precision recall.