计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (15): 44-46.

• 学术探讨 • 上一篇    下一篇

基于改进粒子群算法的支持向量机

周 涛1,2,张艳宁1,袁和金1,邓方安2,陆惠玲3   

  1. 1.西北工业大学 计算机学院,西安 710072
    2.陕西理工学院 数学系,陕西 汉中 723000
    3.陕西理工学院 计算机系,陕西 汉中 723000
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-05-21 发布日期:2007-05-21
  • 通讯作者: 周 涛

SVM based on improvement Particle Swarm Optimization Algorithm

ZHOU Tao1,2,ZHANG Yan-ning1,YUAN He-jin1,DENG Fang-an2,LU Hui-ling3   

  1. 1.School of Computer,Northwestern Polytechnical University,Xi’an 710072,China
    2.Department of Maths,Shaanxi University of Technology,Hanzhong,Shaanxi 723000,China
    3.Department of Computer,Shaanxi University of Technology,Hanzhong,Shaanxi 723000,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-05-21 Published:2007-05-21
  • Contact: ZHOU Tao

摘要: 对求解含线性约束优化问题的粒子群算法(LPSO)进行了改进,给出了应用其训练支持向量机(SVM)的方法。改进后的算法在基本PSO惯性权重策略的基础上加入了基于种群收敛速度的自适应扰动,能够较好地调整算法的全局与局部搜索能力之间的平衡。对双螺旋问题的分类实验表明本文提出的方法稳定性好,训练出的SVM具有较高的分类正确率。

关键词: 支持向量机, 粒子群优化算法, 惯性权重策略

Abstract: Some improvements on the particle swarm optimizer for linearly constrained optimization are put forward.And the method using this improvement algorithm to train Support Vector Machine is presented.An adaptive disturbance based on the population convergence speed is added to the linearly decreasing inertia weight strategy,by which the balance between global and local exploration is adjusted suitably.The experiments on two-spiral problem shows that the algorithm is feasible and robust for support vector machine training.

Key words: Support Vector Machine(SVM), Particle Swarm Optimization(PSO), inertia weight strategy