电子学报

• 学术论文 • 上一篇    下一篇

基于隐反馈的类时齐Markov推荐模型

刘胜宗1, 廖志芳1, 胡佳1, 樊晓平1,2   

  1. 1. 中南大学信息科学与工程学院/中南大学软件学院, 湖南长沙 410075;
    2. 湖南财政经济学院网络化系统研究所, 湖南长沙 410205
  • 收稿日期:2013-09-18 修回日期:2013-12-16 出版日期:2014-04-25
    • 通讯作者:
    • 樊晓平
    • 作者简介:
    • 刘胜宗 男,1986年出生,湖南邵阳人.中南大学信息科学与工程学院博士研究生.研究领域为数据挖掘、推荐系统、智能信息处理.E-mail:lshz179@163.com;廖志芳 女,1968年出生,湖南长沙人,中南大学软件学院副教授、硕士生导师.研究领域为数据挖掘、推荐系统等.
    • 基金资助:
    • 国家科技支撑计划 (No.2012BAH08B00); 国家自然科学基金 (No.61073105); 湖南省自然科学基金 (No.12JJ3074)

Classified Time Homogeneous Markov Model for Recommendation Based on Implicit Feedback

LIU Sheng-zong1, LIAO Zhi-fang1, HU Jia1, FAN Xiao-ping1,2   

  1. 1. School of Information Science and Engineering/School of Software, Central South University, Changsha, Hunan 410075, China;
    2. Laboratory of Networked Systems, Hunan University of Finance and Economics, Changsha, Hunan 410205, China
  • Received:2013-09-18 Revised:2013-12-16 Online:2014-04-25 Published:2014-04-25
    • Supported by:
    • National Key Technology Research and Development Program of the Ministry of Science and Technology (No.2012BAH08B00); National Natural Science Foundation of China (No.61073105); Natural Science Foundation of Hunan Province,  China (No.12JJ3074)

摘要: 传统Markov链模型在用户浏览行为预测方面体现出较好的性能,但不能很好的体现出用户的兴趣度和所推荐的页面的重要性,因此本文提出类时齐Markov模型.该模型给不同的类别用户单独创建时齐Markov模型,并用时齐Markov模型的平稳分布表征用户的访问兴趣和页面的重要程度.本文进而提出了基于隐反馈的类时齐Markov推荐模型,在真实的WEB服务器日志数据上的实验证明,类时齐Markov模型具有更好的推荐性能.

关键词: Web挖掘, 类时齐Markov模型, 平稳分布, 用户聚类, 个性化推荐

Abstract: Markov chain model shows good performance in the user browsing behavior predictions.But it does not work well in reflecting user's interestingness and the importance of the recommended pages.Therefore,this paper proposes classified time homogeneous Markov model.The proposed model create a time homogeneous Markov model separately for every different category of users and use the stationary distribution of the time homogeneous Markov model to characterize users' access interest and pages' importance.Then this paper puts forward a classified time homogeneous Markov model for recommendation based on implicit feedback.The results of experiment with some real WEB server log data show that the proposed model and algorithm have more perfect performance.

Key words: Web mining, classified time homogeneous Markov model, stationary distribution, user clustering, personalized recommendation

中图分类号: