SoRank: incorporating social information into learning to rank models for recommendation

Weilong Yao, Jing He, Guangyan Huang, Yanchun Zhang

Abstract: Most existing learning to rank based recommendation methods only use user-item preferences to rank items, while neglecting social relations among users. In this paper, we propose a novel, effective and efficient model, SoRank, by integrating social information among users into listwise ranking model to improve quality of ranked list of items. In addition, with linear complexity to the number of observed ratings, SoRank is able to scale to very large dataset. Experimental results on publicly available dataset demonstrate the effectiveness of SoRank.