Abstract: This paper proposes a new approach to modeling user navigation sequences based on Web surfing characteristic. The model can efficiently control its scale and is robust to noise by making use of inverse Gaussian distribution describing the depth of user surfing path and Web popularity characteristic. Experiments show that the model can reduce the effect of noise and not only has lower space complexity and higher prediction accuracy, control the network traffic effectively caused by prefetching.

Key words: Web prefetching, PPM, Web surfing characteristic, Adaptive, Anti-noise

摘要： 基于Web浏览特征提出了一种自适应抗噪声的PPM预测模型。模型在构造过程中，利用描述用户浏览深度特征的逆高斯分布及Web流行度特征，对噪声页面及过期数据进行动态移除，分别从纵向和横向上对PPM预测模型规模进行控制。实验表明，该模型对噪声数据的影响有较大的改善，能较好地动态预测用户的Web浏览特征，不仅预测准确率和存储复杂度都有一定程度的提高，而且有效控制了由预取引起的网络流量。

关键词: Web预取, PPM, Web浏览特征, 自适应, 抗噪声