A competition-based recommender system

There are a large number of items with users' ratings on the internet. Instead of predicting items' ratings with the item content or the user-item-rating matrix, this paper proposes an approach to predict items' ratings basing on paired comparisons among contents of items with different ratings. We assume that the user interest on each item can be represented by the combination of different content features, and employ Bradley-Terry model to confirm the user interest value of each feature. Experimental results show that the competition-based recommender system outperforms popular approaches.