Web access path prediction using knowledge discovered from web logs has become an active research area. Web logs provide updated information about the user’s access record to a web site, which contains useful patterns waiting to be discovered and used for improving the web site. In this study, a new approach to web access pattern prediction is proposed. The methodology is based on the case-based reasoning approach, and the main idea is to discover user access patterns by mining the fuzzy association rules from the historical web log data. In our approach, the time duration of each user session is also considered as one of the attributes of a web access case. A fuzzy index tree is used for fast matching of rules. Furthermore, the system s performance is enhanced using the information contained in the user profile through an adaptation process.