融合拓扑势与因子图的在线社交网用户影响力推断

User Influence Inference of Online Social Network Based on Topology Potential and Factor Graph

拓扑势可用于计算在线社交网用户影响力,但是,其对所有用户优化同一个影响参数,导致影响力以相同速度衰减的缺陷尚待改进.此外,拓扑势只适用于节点质量相同的网络.对此,本文融合拓扑势和因子图,提出影响力因子图(Impact Factor Graph,IFG)模型.IFG模型可推断网络用户节点间影响力,并对每个节点分配不同影响参数.维基百科合作编辑数据集上的实证表明,IFG模型可以解决拓扑势在异质网络上的计算问题,并提高用户影响力分析的合理性.

Topology potential can be used to calculate the influence of online social network users.However,it optimizes the same influence parameter for all users,which leads to the influence decayed at the same speed.Moreover,the Topology potential only applies to networks with the same node quality.In this paper,the Impact Factor Graph(IFG)model is proposed by combining topological potential and factor graph.IFG model can infer the influence among network user nodes,and assign different influence parameters to each node.The empirical results on Wikipedia co-edited datasets show that IFG model can solve the problem of topology potential calculation on heterogeneous networks and improve the rationality of user influence analysis.