基于混合Markov模型的用户浏览预测  

Prediction for users' navigation based on hybrid Markov model

作　　者：陈佳[1] 吴军华[1] 

机构地区：[1]南京工业大学信息科学与工程学院,江苏南京210009

出　　处：《计算机工程与设计》2009年第4期903-905,908,共4页Computer Engineering and Design

摘　　要：建立有效的用户浏览预测模型,准确的预测用户的浏览路径,是开发各种Web站点导航工具的关键。传统的Markov模型是一种简单而有效的预测模型,但它存在测准确率低、预测覆盖率低以及存储复杂度高等缺点。通过对传统Markov模型的扩展,并研究了群体用户在Web上的浏览特性,采用了基于混合Markov模型的用户群兴趣导航发现方法。实验结果表明,该方法比传统Markov模型更准确的反映了用户的访问兴趣,可以获得更高的预测准确率与覆盖率,以及有效地降低存储复杂度。Modeling users＇ navigation in the web is the key to build many kinds of tools which can help user navigate the web efficiently. The traditional Markov model is simple and practical, but it gives low prediction accuracy and coverage rate, as well as requires high space complexity. The model for interest navigation pattern discovery based on hybrid Markov model is used by expending traditional Markov model and studying the visiting character of user group. This approach is shown to be superior to traditional Markov model in navigation interest. In particular, it is more accurate in prediction and higher coverage rate, especially, it also have lower space complexity.

关 键 词：预测 WEB导航 用户群 混合马尔可夫模型 兴趣度 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]