Title of Thesis: Using Distinct Information Channels for a Hybrid Web Recommender System Using Distinct Information Channels for a Hybrid Web Recommender System an Excellent Work Environment

Web recommender systems anticipate the information needs of on-line users and provide them with recommendations to facilitate and personalize their navigation. A variety of approaches have been attempted and proposed. Among them, using Web access logs to generate users’ navigational profiles for recommendation is a popular one, given its non-intrusiveness. However, using only one information channel, namely the Web access history, is often insufficient for accurate recommendation prediction. In this thesis, we advocate the use of additional information channels, such as the content of visited pages and the connectivity between Web resources, to better generate the user profile and to build a hybrid recommender system. We test and evaluate our framework with the University of Alberta Computing Science Department Web site data. Our experiments show that this system can significantly improve the quality of Web site recommendation by combining these distinct information channels. In addition, we expand our approach to the context where pages are not content-rich, or content data are absent. We also test our system in an idiosyncratic data set provided by a commercial system – VIVIDESK.