Recommendation strategy using expanded neighbor collaborative filtering
Bin Wang, Qi Gao, Xiaoxue Feng, Feng Pan · 2017
The evaluation of recommender system is often biased towards accuracy, which is hard to balance all participants' interests. In this paper, a novel recommendation strategy using expanded neighbor collaborative filtering (ECF) is presented. Different from the standard collaborative filtering (CF), this recommendation strategy takes into account the second-order neighbors, which are expected to contribute to the coverage and diversity of recommendation. A transferring similarity is proposed to link the given user with second-order neighbors via nearest neighbors. Based on MovieLens dataset, the strategy was test on several typical similarity indexes. The numerical results confirmed the improvements on coverage and diversity compared to the benchmark CF, without affecting accuracy obviously.