Modeling User's Non-Functional Preferences for Personalized Service Ranking

As the number of web services is increasing on the web, selecting the proper web service is becoming a more and more difficult task. How to make the selection results from a list of services more customized towards users’ personal preferences and help users identify the right services for their personal needs becomes especially important under this context. In this thesis, we propose a novel User Modeling approach to generate user profiles on their non-functional preferences on web services, and then apply the generated profiles to the ranking process in order to make personalized selection results. The User Modeling system is based on both implicit and explicit information from the user. Also, this is a flexible model to include different types of non-functional properties. We performed experiments using a real web service dataset with values on various non-functional properties to show the accuracy of our system.