电子学报 ›› 2018, Vol. 46 ›› Issue (11): 2642-2649.DOI: 10.3969/j.issn.0372-2112.2018.11.011

• 学术论文 • 上一篇    下一篇

一种基于证据理论的多类半监督分类算法

盛凯1, 刘忠1, 周德超1, 魏启航2, 冯成旭1   

  1. 1. 海军工程大学兵器工程学院, 湖北武汉 430033;
    2. 66029部队, 内蒙古锡林郭勒盟 011216
  • 收稿日期:2017-12-03 修回日期:2018-04-22 出版日期:2018-11-25
    • 作者简介:
    • 盛凯 男,1991年出生,山东兰陵人,海军工程大学兵器工程学院博士研究生.主要研究方向为轨迹数据挖掘、机器学习及复杂系统建模与仿真.E-mail:shengkai0214@foxmail.com;刘忠 男,1963年出生,山东龙口人,海军工程大学兵器工程学院教授、博士生导师.主要研究方向为系统工程、复杂系统建模与仿真、系统集成技术.E-mail:liuzhong531@yahoo.cn;周德超 男,1972年出生,山东荣成人,海军工程大学兵器工程学院副教授,硕士生导师.主要研究方向为复杂系统建模与仿真、数据挖掘、人工智能.E-mail:13397190531@189.cn
    • 基金资助:
    • 湖北省自然科学基金 (No.2017CFB377)

A Multi-class Semi-Supervised Classification Algorithm Based on Evidence Theory

SHENG Kai1, LIU Zhong1, ZHOU De-chao1, WEI Qi-hang2, FENG Cheng-xu1   

  1. 1.College of Weapons Engineering, Naval University of Engineering, Wuhan, Hubei 430033, China;
    2.PLA 66029 Troop, Xilinguole, Inner-Mongolia 011216, China
  • Received:2017-12-03 Revised:2018-04-22 Online:2018-11-25 Published:2018-11-25
    • Supported by:
    • Natural Science Foundation of Hubei Province,  China (No.2017CFB377)

摘要: 为了提高多类半监督分类的性能,提出了一种基于证据理论的多类协同森林算法(DSM-Co-Forest).首先,通过"多对多"模式将有标记的多类数据随机拆分为多个二类数据集,并以此训练二类基分类器;然后,利用多个基分类器同时对未标记样本进行预测,并利用证据组合算法挑选出可信度较高的未标记样本;最后,将高可信度的未标记样本加入到原训练样本中,以迭代更新其他的基分类器,从而提高分类器的整体性能.通过在一些公共数据集上进行实验,并与其他半监督分类算法进行对比,验证了所提算法的可行性和有效性.

关键词: 半监督学习, 多类分类, 证据理论, 协同森林

Abstract: In order to improve the performance of multi-class semi-supervised classification, a new multi-class Co-Forest algorithm named DSM-Co-Forest is proposed on the basis of D-S evidence theory. First, through MVM mode, the multi-labeled data set is randomly split into multiple binary-class data set to train the base classifiers; then, these base classifiers are used to pick out the high reliability samples from the unlabeled data set by using the evidence combination algorithm; finally, adds these selected samples to the original training set to iteratively update the base classifiers so as to improve the overall performance of the multi-class classifier. Through comparing with other semi-supervised classification algorithms on several public data sets, the feasibility and validity of the proposed algorithm are verified.

Key words: semi-supervised learning, multi-class classification, evidence theory, co-forest

中图分类号: