用户浏览页面预测的多阶HMM模型融合

Multiple HMM Fuzzy Integral Algorithm for Web Access Prediction

预测用户的浏览页面是WWW上的一个重要的研究方向.提出了一种通过模糊积分融合多阶HMM（Hidden Markov Model）预测结果的用户浏览页面预测模型.该算法首先拓展了经典融合预测算法的先验信息空间.其方法是通过对不同用户浏览模式分类,建立1阶多Morkov链模型并以其训练结果为权重指标,而后通过模糊积分理论融合1~N阶HMM预测的结果.性能测试实验结果表明该模型预测准确率优于已有的用户浏览页面预测的多HMM融合方法.该方法可在Web站点管理、电子商务以及网页预取等领域广泛应用.

Web-user access prediction has recently attracted significant attention.In this paper,we focus on predicting Web access and present a new fuzzy multiple HMM model integral algorithm.It extends the prior information of traditional fusion models and introduces Fuzzy integral algorithm into combine the prediction of 1～N order HMM models.The experimental results show that our model can provide obvious improvements over previously proposed Web access predicting algorithms.The method can be widely used in Web site management,E-commerce and Web prefetching,etc.