An Adaptive PPM Prediction Model Based on Pruning Technique

The key issue of Web prefetching is to establish an effective user prediction model. Prediction by partial match (PPM) is one of the context models used in the Web prefetching area. The high space complexity and low efficiency of the PPM model affect its application. In this paper, we make use of pruning technique and propose a new adaptive PPM model based on Zipf´s law and Web access characteristics. The experiments have shown that this model not only can be used to make predictions dynamically, but also has relative lower space complexity and higher prediction accuracy.