REXWERE: A TOOL FOR FUZZY RULE EXTRACTION IN WEB RECOMMENDATION

In this paper, we present REXWERE, a software tool designed and implemented in order to extract knowledge from Web usage data in the form of recommendation fuzzy rules useful to provide personalized link suggestions to the visitor of a Web site. REXWERE employs a hybrid approach that combines fuzzy reasoning and neural learning within a working scheme made of several steps. Firstly, a fuzzy clustering process is applied to group similar user sessions into user profiles. Next, a neuro-fuzzy network is trained using information about user profiles in order to derive a set of recommendation fuzzy rules. Finally, a further learning step is performed to improve the accuracy of the derived recommendation model. Throughout the use of REXWERE, the user is guided by a wizard-based interface made of a sequence of panels. Each panel consists in a graphical window providing a basic function of the tool. An illustrative example is provided to show the use of REXWERE and to demonstrate its effectiveness in finding good recommendation rules.