A real-time evolutionary algorithm for Web prediction

As an increasing number of users access information on the World Wide Web, there is a opportunity to improve well known strategies for Web caching, prefetching, dynamic user modeling and dynamic site customization in order to obtain better subjective performance and satisfaction in Web surfing. We propose a new method to exploit user navigational path behavior to predict, in real-time, future requests. Predicting user next requests is useful not only for document caching/prefetching, it is also suitable for quick dynamic portal adaptation to user behavior. Real-time user adaptation prevents the use of statistical techniques on Web logs, and we propose the adoption of a predictive user model based on finite state machines together with an evolutionary algorithm that evolves a population of FSMs for achieving a good prediction rate.