The Feature Selection Method for SVM with Discrete Particle Swarm Optimization Algorithm

摘要：

Aiming at solving the problem of feature selection in support vector machine (SVM) training process, a new feature selection method for SVM based on discrete particle swarm algorithm is proposed. The new method employs the support vector rate as the evaluation criterion for the performance of SVM and directs the search of the particle swarm algorithm based on this. The UCI benchmark data sets experiments demonstrate that the new method can find the best feature subset for SVM and improve the classification accuracy and running efficiency greatly.

关键词： support vector machine particle swarm optimization feature selection

作者: PENG Xiyuan WU Hongxing PENG Yu

作者单位: Dept of Automatic Test and Control Harbin Institute of Technology, Science Park of Harbin Institute of Technology, P.O.Box 3033,Harbin Heilongjiang China, 150080

会议类型: 国际会议

会议名称: 第七届国际测试技术研讨会

会议地点: 北京

会议语种:英文

在线出版日期: 2007-08-05（万方平台首次上网日期，不代表论文的发表时间）