User preference boosted content-based recommender system

Özberk Yener, Tuğçe
In the world of information, internet becomes the most important information source. However, internet contains vast amount of information and this information is not filtered. In such an environment, the people who seek for an information is overwhelmed in the alternatives that s/he can reach via the web. Recommender systems have their real importance in this kind of situations. To overcome said overwhelming problems, recommender systems are developed to determine the people needs and to recommend suitable alternatives to them. The current recommendation methods are classified under three main categories: collaborative filtering, content-based and hybrid approaches. Classical content-based recommendation approaches include the content information of the items. In this thesis work, we propose a user preference boosted content-based recommendation methodology. In addition to the items content information, we aimed to define a novel approach to the problem of including user’s preference information to the information filtering process. The novel solution that we explained in this thesis work uses the users past like and dislike rate information related to the specific items to predict recommendation scores related to the unseen items. The results which we obtained by implementing the proposed user preference boosted content based recommendation approach indicates that; by including the users' preference information to the items content information more accurate recommendations can be done and more reliable results can be gathered. We present the implementation details and comparative evaluation results of the proposed novel approach in this thesis.
Citation Formats
T. Özberk Yener, “User preference boosted content-based recommender system,” M.S. - Master of Science, Middle East Technical University, 2013.