Skewness-aware Clustering Social Recommender

Recommender systems have been a hot research area recently. One of the most widely used methods is Collaborative Filtering(CF), which selects items for an individual user from other similar users. However, CF may not fully reflect the procedure of how people choose an item in real life, for users are more likely to ask friends for opinions instead of asking strangers with similar interests. Recently, some recommendation methods based on social network have been proposed. These approaches assume that data follow Gaussian distribution, and then incorporate social network into the CF algorithms. Therefore, users preferences can be influenced by the flavors of their friends. In this paper, we propose the Skewness-aware Clustering Social Recommender. Our contributions are three-fold: (1) we show that data do not follow Gaussian distribution and then correct the bias and skewness of ratings; (2) we propose the social cluster model based on cluster coefficient; (3) our improvement outperforms the baseline by 3.34%.