Abstract

Web is a large network of related documents that are increasing rapidly. Due to the unknown optimal structure for a web site, providing a way to do this is important. In this paper, a method based on distributed learning automaton (DLA) with variable number of actions to create an adaptive web site is provided that has ability to insert efficient links and eliminate inefficient links. In the proposed method, we used web usage data that had tried to change the structure of the web site based on users’ interaction in the past. Simulation results show that in the new structure, users can achieve their desired information with less number of steps than primary web site and the proposed method because considering the variable learning parameter has a higher efficiency than existing algorithms and does not need to readjust the learning parameters if numbers of web pages increase.

Key words: Web usage mining, adaptive web sites, site improvement, site reorganization, learning automata