Skip to Main Content
In this paper we study data mining techniques as tools for reducing the time needed to access Web pages in the environment of a corporation where users connect to the Internet through a proxy server. We add a data mining server to the traditional Web architecture. This server computes, using sequential patterns, the pages likely to be requested by the users, taking into account their Web access history. Then, the server loads these pages into the proxy's cache in order to have them available when they are actually asked for. We describe our implementation and the results obtained, showing that the average access time to a Web page can be dramatically reduced using this technique. Further, we also implemented our proposal using simple association rules, showing that sequential patterns result in a smaller and more accurate set of rules, as a consequence of taking into account the order of the requests. Finally, we discuss the differences with other prefetching proposals.