Relational fuzzy approach for mining user profiles

Capturing the characteristics and preferences of Web users into user profiles is a fundamental task to perform in order to implement forms of personalization on a Web site. In this paper, we present a relational fuzzy clustering approach to extract significant user profiles from session data derived from log files. In particular, a modified version of the CARD clustering algorithm is proposed in order to produce well distinct clusters corresponding to profiles reflecting the actual user preferences embedded in the available session data. Experimental results on session data extracted from log files of a sample Web site are reported.