计算机与现代化 ›› 2010, Vol. 1 ›› Issue (02): 176-179.doi:

• 应用与开发 • 上一篇    下一篇

基于Cotraining的烟草原料数据优化分析

卢加磊1,朱世华2,丁香乾3,黄跃华2   

  1. 1.中国海洋大学信息科学与工程学院,山东 青岛 266071; 2. 红塔烟草(集团)有限责任公司,云南 玉溪 653100; 3.中国海洋大学信息工程中心,山东 青岛 266071
  • 收稿日期:2009-07-08 修回日期:1900-01-01 出版日期:2010-02-12 发布日期:2010-02-12

Optimized Analysis of Tobacco Materials Data Based on Cotraining

LU Jialei1,ZHU Shihua2,DING Xiangqian3,HUANG Yuehua2   

  1. 1.Department of Information and Engineering, Ocean University of China, Qingdao 266071, China; 2. Hongta Tobacco (Group) Co., Ltd., Yuxi 653100, China;3.Information and Engineering Center, Ocean University of China, Qingdao 266071, China
  • Received:2009-07-08 Revised:1900-01-01 Online:2010-02-12 Published:2010-02-12

摘要:

本文根据烟草行业对原料数据综合分析的实际需要,结合机器智能学科中半监督体系内的Cotraining方法进行理论和应用分析。本文在理论分析的基础上得到Cotraining方法应用于烟草原料数据优化分析的机器学习模型,并且通过实验数据的总结验证和与其他算法的比较,表明此算法模型具有一定优越性的结果。

关键词: 半监督学习, 协同训练, 期望最大化, 朴素贝叶斯算法, 烟草数据分析

Abstract:

This thesis combins the two important studies machine intelligence and tabacco industry to solve insdustial material analysing research subjects. The author explains the theoretical and applied analysis of the cotraining algorithm which belongs to semisupervised learning framwork. Based on analysis of the theroy and need of tobacoo industy, the author constructs a machine learning model based on cotrainging which apply to optimized analysis of tobacco materials data. Some experiments are done to verify the correctness and superiority compared to other methods.
[WTHZ]Key words:[WTBZ]

Key words: semisupervised learning, cotraining, EM, Nave Bayes, analysis of tabacco materials