计算机工程与应用 ›› 2015, Vol. 51 ›› Issue (13): 139-144.

• 数据库、数据挖掘、机器学习 • 上一篇    下一篇

改进蚁群算法在SVM参数优化研究中的应用

高雷阜,张秀丽,王  飞   

  1. 辽宁工程技术大学 理学院,辽宁 阜新 123000
  • 出版日期:2015-07-01 发布日期:2015-06-30

Application of improved ant colony algorithm in SVM parameter optimization selection

GAO Leifu, ZHANG Xiuli, WANG Fei   

  1. College of Science, Liaoning Technical University, Fuxin, Liaoning 123000, China
  • Online:2015-07-01 Published:2015-06-30

摘要: 支持向量机参数的选择决定着支持向量机的分类精度和泛化能力,而其参数优化缺乏理论指导,在此背景下提出了ACO-SVM模型。该模型将SVM分类预测准确率作为目标函数,对蚁群算法进行改进,引入有向搜索和基于时变函数更新的信息素更新原则,利用蚁群算法的并行性、正反馈机制和较强的鲁棒性,以求得最优目标并得到SVM的最优参数组合。数值实验结果表明,改进蚁群算法在SVM参数优化选取中具有更好的寻优性能,具有较高的分类准确率;该方法具有较好的并行性和较强的全局寻优能力。

关键词: 支持向量机, 蚁群优化算法, 参数优化, 分类正确率

Abstract: SVM parameter selection determines SVM classification accuracy and generalization ability, and its lack of theoretical guidance parameter optimization, ACO-SVM model is proposed, it predicts the SVM classification accuracy as the objective function, and improves the ant colony algorithm, with the introduction of search and updates the pheromone based on time-varying function update policy, uses the ant colony algorithm parallelism, positive feedback mechanism and strong robustness, in order to achieve optimal goals and get the optimal combination of parameters of SVM. The results of numerical value experiments show that the improved Ant Colony Optimization algorithm for SVM parameters selection has better optimization performance and higher classification accuracy. This method has the better parallelism and strong global optimization ability.

Key words: Support Vector Machine(SVM), Ant Colony Optimization Algorithm(ACOA), parameter optimization, classification accuracy