The adaptability of a web application is its ability to react depending on the needs and the preferences of users. Thus, user models, used by such adaptive user interface, contain personal information which is required for learning personalized process. Then, evaluation of web applications interests on how users can learn to achieve their objectives. To gather this information a variety of measures have been used. In this paper, we investigate and present our adaptive Web interface using a Bayesian networks approach and we give a special importance to the evaluation of this web interface. The experiments show that the adaptive Web interface provides results that satisfy the user. We confirmed that the adaptive user interface was more comfortable for use than the fixed user interface.