基于滑动窗口的自适应网页预测模型

Sliding Window-Based Adaptive Web Prediction Modeling

PPM模型广泛应用于Web预取技术,但大多数的PPM模型不具有自适应性,不能反映用户浏览模式的改变。通过对标准PPM模型的扩展,提出基于滑动窗口的自适应网页预测模型。该模型仅保留处于滑动窗口之内的最近访问序列,从而反映用户兴趣的变化,同时利用非压缩后缀树增量式添加新的用户请求和删除过时的浏览信息,以提高更新速度。实验表明,该模型能更准确地描述用户在Web上的浏览特征,在预取性能上明显地优于以往的模型。

Prediction by partial match （PPM） models are commonly used for web prefetching. But most of existing models are not adaptive and can not represent the change of user browsing behaviors. By extending the standard PPM model, we present an adaptive Web prediction model based on sliding window. The model only keeps the most recent requests by a sliding window to indicate user interest changing. In order to improve the updating speed, it makes use of non-compact suffix tree to incrementally insert the new user request and delete the outdated browsing information. Trace-driven experiments show that our model can significantly improve the prefetching performance.