Computer Engineering and Applications ›› 2015, Vol. 51 ›› Issue (4): 110-114.

Previous Articles     Next Articles

Application of GA_SJ in SVM parameter optimization

GAO Leifu, ZHANG Xiuli, TONG Pan   

  1. College of Science, Liaoning Technical University, Fuxin, Liaoning 123000, China
  • Online:2015-02-15 Published:2015-02-04

GA_SJ在SVM核参数优化中的应用

高雷阜,张秀丽,佟  盼   

  1. 辽宁工程技术大学 理学院,辽宁 阜新 123000

Abstract: Support Vector Machine(SVM) is a kind of machine learning methods which has good performance, but there still lacks the system theory as a guide for the choice of its parameters. Aimming at the shortage of the classical SVM parameter selection method, genetic algorithm, this paper comes up with the improved algorithm, and combines it with the SVM, gets the algorithm namely GA_SJ algorithm, which can choose the kernel parameters and train SVM automatically. This algorithm introduces the random search into genetic algorithm, and adopts the optimal preservation strategy and dynamic crossover and mutation probabilities, and greatly improves the efficiency of genetic algorithm. The numerical experiment results confirm that GA_SJ algorithm is feasible and effective in the SVM parameter optimization, and the SVM obtained has higher classification performance.

Key words: Support Vector Machine, genetic algorithm, random search, parameter optimization

摘要: 支持向量机(SVM)是一种性能良好的机器学习方法,但是对于其参数的选择还缺少系统的理论作为指导。针对经典的SVM参数选择方法——遗传算法的一些不足,提出了改进,并将其与SVM相结合,得到自动选择核参数并进行SVM训练的算法即GA_SJ算法。该算法通过将随机搜索引入到遗传算法当中,并采用最优保存策略和动态的交叉和变异概率,有效地提高了遗传算法的效率。数值实验结果证实了GA_SJ算法在SVM参数优化中的可行性和有效性,而且得到的SVM具有较高的分类性能。

关键词: 支持向量机, 遗传算法, 随机搜索, 参数优化