Intelligent web site is a new portal generation, able to improve its structure and content based on the analysis of the user behavior. This paper focuses on modeling the visitor behavior, assuming that the only source available is his/her browsing behavior. A framework to acquire and maintain knowledge extracted from web data is introduced. This framework allows to give online recommendations about the navigation steps, as well as offline recommendations for changing the structure and contents of the web site. The proposed methodology is applied to the web site of a commercial bank.