计算机工程 ›› 2012, Vol. 38 ›› Issue (7): 174-176.doi: 10.3969/j.issn.1000-3428.2012.07.057

• 人工智能及识别技术 • 上一篇    下一篇

基于PSO拓展的多分类器加权集成方法

刘擎超1,朱玉全1,陈 耿2   

  1. (1. 江苏大学计算机科学与通信工程学院,江苏 镇江 212013;2. 南京审计学院信息科学学院,南京 211815)
  • 收稿日期:2011-08-18 出版日期:2012-04-05 发布日期:2012-04-05
  • 作者简介:刘擎超(1987-),男,硕士研究生,主研方向:机器学习,数据挖掘;朱玉全,教授、博士生导师;陈 耿,教授、博士
  • 基金项目:
    国家自然科学基金资助项目(70971067);江苏省自然科学基金资助项目(BK2010331)

Multiple Classifiers Weighted Integration Method Based on Particle Swarm Optimization Expanded

LIU Qing-chao 1, ZHU Yu-quan 1, CHEN Geng 2   

  1. (1. School of Computer Science and Telecommunications Engineering, Jiangsu University, Zhenjiang 212013, China 2. School of Information Science, Nanjing Audit University, Nanjing 211815, China)
  • Received:2011-08-18 Online:2012-04-05 Published:2012-04-05

摘要: 在多分类器集成时,每个基分类器的效能不同,如每个权值都相同,则会影响基分类器发挥作用。基于此,提出基于PSO拓展的多分类器加权集成方法BCPSO。该方法采用随机子空间生成各个独立的子分类器,输出结果通过各分类器加权投票组合规则集成。实验结果表明,该方法有效可行,具有较高的分类正确率。

关键词: 基分类器, 加权投票, 分类器, 随机子空间, 粒子群优化

Abstract: For the integration of multiple classifiers, each base classifier performance is different, if the weights are the same for each classifier, it is bound to affect the classification of base classifier. This paper proposes a weighting method Binary Crossover Particle Swarm Optimization (BCPSO), in which each individual sub-classifier uses random subspace method to generate and output the final classification by the combination of rule by weighted voting. Experimental results show that the method is feasible and effective, it has high classification accuracy.

Key words: base classifier, weighted voting, classifier, random subspace, Particle Swarm Optimization(PSO)

中图分类号: