A novel approach to generate artificial outliers for support vector data description

In this paper, we propose a novel approach to generate artificial outliers for support vector data description with boundary value method. In SVDD, the width parameter s and the penalty parameter C influence the learning results. The N-fold M times cross-validation is well-known and popular scheme to calculate the best (C , s ) values. To automatically optimize the identification rate, we need more outliers. Due to this reason, we utilize boundary value in any two dimensions randomly to generalize new outliers. At the last, we use three benchmark data sets: Iris, Wine, and Balance-scale data base to validate the approach in this research has better classification result and faster performance.

[1]  Hwanjo Yu.  SVMC: Single-Class Classification With Support Vector Machines , 2003, IJCAI.

[2]  L. Zhuang,et al.  Parameter optimization of Kernel-based one-class classifier on imbalance text learning , 2006 .

[3]  Chi-Kai Wang,et al.  A novel approach of feature classification using Support Vector Data Description combined with interpolation method , 2008, 2008 34th Annual Conference of IEEE Industrial Electronics.

[4]  Robert P. W. Duin,et al.  Support Vector Data Description , 2004, Machine Learning.

[5]  Robert P. W. Duin,et al.  Uniform Object Generation for Optimizing One-class Classifiers , 2002, J. Mach. Learn. Res..

[6]  Xuan Wang,et al.  Sphere Classification for Ambiguous Data , 2006, 2006 International Conference on Machine Learning and Cybernetics.

[7]  Robert P. W. Duin,et al.  Support vector domain description , 1999, Pattern Recognit. Lett..

[8]  Xue-Qin Shen,et al.  A new method for edge detection based on support vector classification , 2002, Proceedings. International Conference on Machine Learning and Cybernetics.

[9]  Vladimir Naumovich Vapni.  The Nature of Statistical Learning Theory , 1995 .

[10]  Vladimir N. Vapnik,et al.  The Nature of Statistical Learning Theory , 2000, Statistics for Engineering and Information Science.

[11]  M. M. Moya,et al.  One-class classifier networks for target recognition applications , 1993 .