作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (21): 171-174. doi: 10.3969/j.issn.1000-3428.2012.21.046

• 人工智能及识别技术 • 上一篇    下一篇

基于混合多距离度量的多分类器加权集成研究

赵玉娟,刘擎超   

  1. (江苏大学计算机科学与通信工程学院,江苏 镇江 212013)
  • 收稿日期:2011-12-28 出版日期:2012-11-05 发布日期:2012-11-02
  • 作者简介:赵玉娟(1987-),女,硕士研究生,主研方向:智能交通,数据挖掘,车载自组网;刘擎超,硕士研究生

Research on Weighted Integration of Multiple Classifier Based on Hybrid Multiple Distance Measurement

ZHAO Yu-juan, LIU Qing-chao   

  1. (School of Computer Science and Telecommunication Engineering, Jiangsu University, Zhenjiang 212013, China)
  • Received:2011-12-28 Online:2012-11-05 Published:2012-11-02

摘要: 在机器学习领域,分类器加权在小样本数据集中的分类正确率较低。为此,提出一种基于混合距离度量的多分类器加权集成方法。结合欧氏距离、曼哈顿距离、切比雪夫距离,设计混合的距离度量加权方法,使用加权投票组合规则集成各分类器的输出结果。实验结果表明,该方法鲁棒性较好,分类正确率较高。

关键词: 欧氏距离, 度量方法, 混合多距离, 分类器, 加权集成

Abstract: Aiming at the problem that the low classification accuracy rate of small sample in machine learning, this paper proposes a weighted integration of multiple classifiers method based on Hybrid Multiple Distances(HMD) measurement. In this method, three measure methods are combined, and the output of each classifier is integrated by weighted majority voting. It designs the mixed distance measure weighted method, and uses the weighted voting combination rules to integrate the classifier output results. Experimental results show that this method has good robustness and high classification accuracy.

Key words: Euclidean distance, measurement method, Hybrid Multiple Distance(HMD), classifier, weighted integration

中图分类号: