基于模体演化与社区一致性的时序链路预测方法

Time series link prediction method based on motif evolution and community consistency

针对动态有向网络中的时序链路预测问题,充分分析动态有向网络中微观结构三元组模体的演化规律,使用指数平滑法季节加法(Holter-Winter-Additive)时序分析方法预测三元组模体的转换概率,引入牛顿法寻求时序分析方法中的最优参数;同时考虑到节点的社区属性对链路预测产生的影响,定义模体内节点的社区结构一致性重要指标,对三元组模体的影响力进行评估。基于此,首先使用时间序列分析方法对模体的转换概率进行预测,进而结合模体社区结构一致性的指标提出一种新的链路预测方法。使用不同的方法在三个真实的有向网络中进行验证,实验结果显示该方法能够达到更好的链路预测效果。

In order to solve the problem of time-series link prediction in dynamic directed network,this paper investigated the evolution law of triple motifs over time,and proposed the method to predict the transfer probability of motif based on the time series analysis model of Holter-Winter-Additive. Then it introduced a Newton method to optimize the parameters. At the same time,this algorithm considered the impact of the community attribute of node on link prediction. It also defined the index of community consistency to evaluate the influence of the triad motif. On the basis of this analysis,it adopted the time series analysis model to extract the evolution regulars of motif,and proposed a new temporal link prediction method combining with the community consistency of motif. Lots of comparative experiments have been done in three real networks. The results demonstrate that this approach can effectively improve the link prediction accuracy.