Fusion-based recommender system

Recommender systems have become a common way to help people navigate among increasingly more complex selection options. Various recommendation algorithms have been proposed and a lot of business solutions have been built for applications such as selecting books, cinema, video/TV program, restaurants, etc. However, there is no simply best approach due to the high complexity and uncertainty of the problem. Therefore, a recommender system usually includes multiple different recommenders and aggregates their recommendation results through a combiner. In this paper, we discussed combining recommenders in the framework of information fusion theory including rank fusion, decision fusion, Dempster-Shafer fusion and estimation fusion. Experiments results on the benchmark Movie-Lens dataset show that our proposed methods with fusion techniques leaded to significant performance improvements over the baselines models.