Abstract—For accelerating the training speed of Support Vector Machine (SVM) when solving large scale unbalanced learning problem, a novel parallel incremental proximal linear SVM was proposed. This paper proposed a new strategy to obtain the weight matrix to overcome the unbalanced property, thus acquired higher prediction accuracy for unbalanced dataset. Moreover, a cascade architecture of parallel SVM methods was used to accelerate the training speed. Experiments show that this method could not only improve prediction accuracy, but also increase the training speed and save more training time.

Index Terms—Incremental algorithm, Parallel, Support Vector Machine, Unbalanced Dataset

Xin Zhang is with the Department of Electronic Engineering, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong SAR. (corresponding author to provide phone: 0852-3442-3791; e-mail: marix4@gmail.com).
GengFeng Zhu was with College of Information Science and Engineering, Shandong University of Science and Technology, Shandong 266510, China. (e-mail: zwolfe@163.com).

Cite: Xin Zhang and Gengfeng Zhu, "Parallel Incremental Proximal Linear Support Vector Machine," International Journal of Computer and Electrical Engineering vol. 2, no. 2, pp. 292-295, 2010.