Web Object Prefetching: Approaches and a New Algorithm

In Internet applications, Web object prefetching is a commonly used and quite effective algorithmic approach to reduce user perceived delays. While a separate concept, prefetching is closely related to caching and they are often blended together in Web algorithms. In this paper, we give a review of Web prefetching models and algorithms, categorize them into groups based on their design principles, and compare their functionalities and performance. We then proposed a new prefetching algorithm that is based on the Adaptive Resonance Theory (ART) of neural networks. The new model uses the bottom-up and top-down weights of the cluster-URL connections obtained from a modified ART1 algorithm to make prefecthing decisions.