吉林大学学报(理学版)

• 计算机科学 • 上一篇    下一篇

基于最大差距的置信度评估算法

董立岩1, 朱琪1, 隋鹏1, 孙鹏1, 李永丽2   

  1. 1. 吉林大学 计算机科学与技术学院, 长春 130012; 2. 东北师范大学 计算机科学与信息技术学院, 长春 130117
  • 收稿日期:2015-07-02 出版日期:2015-11-26 发布日期:2015-11-23
  • 通讯作者: 朱琪 E-mail:zhuqi13@mails.jlu.edu.cn

Confidence Evaluation Algorithm Based on the Maximum Distinction

DONG Liyan1, ZHU Qi1, SUI Peng1, SUN Peng1, LI Yongli2   

  1. 1. College of Computer Science and Technology, Jilin University, Changchun 130012, China; 2. School ofComputer Science and Information Technology, Northeast Normal University, Changchun 130117, China
  • Received:2015-07-02 Online:2015-11-26 Published:2015-11-23
  • Contact: ZHU Qi E-mail:zhuqi13@mails.jlu.edu.cn

摘要:

估方法, 并在UCI数据集上对两种方法进行对比实验. 实验结果表明, 基于最大差距的置信度评估方法在宏平均召回率、 宏平均精度及所用时间上均优于K邻近置信度评估方法, 从而可进一步优化半监督分类学习中数据样本的置信度评估.

关键词: 置信度评估, 分类, 半监督学习, K-邻近算法

Abstract:

The authors proposed two confidence estimating methods, namely, K-nearest and maximum distinction methods, which were practized with selected certain percentage of data in the UCI datasets, compared macrorecall, macroprecision and time loss, and proved the effectiveness of maximum distinction. The results show the improvement of confidence evaluation in semisupervised classification.

Key words: confidence evaluation, classification, semisupervised learning, K-nearest algorithm

中图分类号: 

  • TP301.6