OmniRank: learning to recommend based on omni-traversal of heterogeneous graphs
Social Network Analysis and Mining(2019)
摘要
In this paper, we propose a new node similarity measure, OmniRank, for multi-dimensional and heterogeneous social networks. In particular, we recursively propagate the structural similarity computation beyond the neighborhood of the nodes to the entire heterogeneous (e.g., user, item, tag) graph, which incorporates several unipartite and bipartite graphs. We have evaluated experimentally OmniRank and compared it against other state-of-the-art algorithms (wRWR, SimRank and P-Rank) on two real-life data sets (HetRec 2011 and GeoSocialRec). Our experiments have shown that OmniRank outperforms its comparison partners in terms of effectiveness and recommendation accuracy, because it exploits information on both multi-step and omni-directional neighborhoods (unipartite and bipartite).
更多查看译文
关键词
Algorithms, Link prediction, Product recommendation, Social networks
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络