Acquiring knowledge about user's preferences in a Web site

Acquiring knowledge about user preferences in a Web site can be the key for continuous improvement of the site. A way to obtain it consists in understanding the user's browsing behavior, analyzing the Web log files with data mining techniques. From the set of possible variables that contain information on the visitors behavior, two have been selected: the time spent on each visited page and the respective page content which is based on the importance of each word in a page. We propose an importance measure that considers additionally if visitors search for a specific word. With this information, a new similarity between user sessions is defined. This similarity can be applied in any clustering algorithm in order to cluster similar user sessions. Using a self-organizing feature map we cluster sessions on a certain Web site. The respective results gave very important insights regarding visitors behavior and preferences and prompted the reconfiguration of the Web site.