Televizyon programları için çizge tabanlı işbirlikçi ve bağlam duyarlı öneri sistemi.

With the increasing amount of TV programs and the integration of broadcasting and the Internet with smart TV’s, users suffer the difficulty of selecting the most appealing TV programs among various different programs available. User decisions are mostly affected by the contextual properties of programs such as the time of day, genre, actors and directors of program. This thesis proposes the design, development and evaluation of a graph based context-aware collaborative recommender system for TV programs. The proposed graph based algorithm is based on random walks performed on a tri-partite graph. The graph is constructed by using context aware pre-filtering in order to filter out programs which are irrelevant in the given context. The recommendation list generated by the graph based collaborative algorithm is updated by re-ranking the recommended items according to additional contextual variables. Thus, the proposed recommender system exploits both contextual pre- filtering and post-filtering to produce more effective recommendations. In order to measure the effectiveness of the context variables, we have implemented evaluation metrics on both context-free and contextual graph based methods. We have also tested the effects of parameters used in the graph based collaborative algorithm to the success of the recommender. The results indicate that context can provide better recommendations for TV programs. .M.S. - Master of Scienc