计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (8): 120-125.DOI: 10.3778/j.issn.1002-8331.1510-0150

• 模式识别与人工智能 • 上一篇    下一篇

改进的MEA进行特征选择及SVM参数同步优化

丁  胜1,2,3,张  进2,3,李  波2   

  1. 1.武汉大学 电子信息学院,武汉 430080
    2.武汉科技大学 计算机科学与技术学院,武汉 430065
    3.武汉科技大学 智能媒体计算湖北省重点实验室,武汉 430065
  • 出版日期:2017-04-15 发布日期:2017-04-28

Improved MEA for feature selection and SVM parameters optimization

DING Sheng1,2,3, ZHANG Jin2,3, LI Bo2   

  1. 1.School of Electronic Information, Wuhan University, Wuhan 430080, China
    2.School of Computer Science & Technology, Wuhan University of Science and Technology, Wuhan 430065, China
    3.Hubei Province Key Laboratory of Intelligent Media Calculation, Wuhan University of Science and Technology, Wuhan 430065, China
  • Online:2017-04-15 Published:2017-04-28

摘要: 特征选择和参数优化是提高支持向量机(SVM)分类性能的两个重要手段,将两者进行同步优化能提高分类器的分类精度。利用思维进化算法(MEA)进行特征选择和SVM参数同步优化能取得较好的分类效果,但也存在着收敛速度慢,易陷入局部最优的问题,无法进一步提高分类精度。针对这一问题,提出了一种改进的思维进化算法进行分类器优化(RMEA-SVM),在传统思维进化算法的基础上引入了“学习”和“反思”机制,利用子群体间信息共享进行学习,通过适应度值的比较进行反思。通过这种方式保证种群的多样性,加快收敛速度,进一步提高分类精度。实验结果证明了算法的有效性。

关键词: 支持向量机, 特征选择, 参数优化, 思维进化算法, 学习, 反思

Abstract: Feature selection and parameter optimization are two main methods which improve the classification performance of Support Vector Machine(SVM). The optimization of feature selection and parameter synchronization can improve the classification accuracy. Mind Evolutionary Algorithm(MEA) is used for feature selection and SVM parameters optimization can achieve good classification results, but there are also some problems such as slow convergence speed and easy to fall into local optimum which lead to the classification accuracy could not be further improved. Aiming at this problem, an improved mind evolutionary algorithm for feature selection and SVM parameter optimization is proposed into which the “learning” and “reflection” mechanism are introduced. The algorithm learns through the information sharing among subgroups and reflects through the comparison of the fitness value. In this way it ensures the diversity of population and increases the speed of convergence, and further improves the classification accuracy. Experimental results show the effectiveness of the algorithm.

Key words: support vector machine, feature selection, parameter optimization, mind evolutionary algorithm, learning, reflection