The thesis describes different types of collaborative filtering methods to filter information from the large amount available and presents examples of such systems in different domains. It focuses on automated collaborative filtering to generate personalized recommendation of information.
Different variations of the automated collaborative filtering scheme are developed and analyzed in the thesis. An additional adjustment of the predicted score is implemented in order to improve precision of the recommendation. Different combinations of parameters are analyzed to maximize system effectiveness.
The data for the analysis was gathered through TV Recommender, a World Wide Web system developed for the thesis. The TV Recommender is a fully functional system that acquires users' data and implements the enhanced collaborative filtering scheme to generate user's personalized TV recommendation.
|