Predicting web accesses using personal history

Web resources are often cached by clients. Caching allows bandwidth preservation since resources that have already been obtained could be reused. However, it is possible only for those resources that have been requested in the past. Prefetching is a technique to fetch and cache resources that have never been obtained through normal usage. If prefetching a resource overlaps with the use of another resource, the subsequent use of the prefetched resource would appear to have less latency compared to normally fetching the resource on demand. The core of prefetching is prediction - predicting which resources might be used in the near future. Most of the published work on web prefetching is focussed on prediction algorithms that operate on server-side data. These algorithms attempt mask the access latency perceived by all the clients connecting to the server. Such an approach may not necessarily mask the access latency perceived by an individual client. This paper focusses on individual clients and examines how client-side prediction using personal history can improve the latency perceived by an individual user.