基于节点嵌入的权重符号社交网络高效链路预测算法

AN EFFICIENT LINK PREDICTION ALGORITHM BASED ON NODE EMBEDDING IN WEIGHTED SIGNED SOCIAL NETWORKS

目前大部分链路预测的相关研究均集中在无符号/无权重的社交网络,对于WSSN中的链路预测问题,特别是边权重预测问题的研究较少。提出一种高效的权重符号社交网络嵌入(Weighted Signed Network Embedding,WSNE)算法,对每一条有向边,学习其两个端点的嵌入表示,同时考虑边符号约束来提高权重预测的性能。在三个真实WSSN数据集上与其他相关算法进行比较,实验结果表明,该算法可以有效地预测边权重,并且优于相关算法。

At present,most relevant researches on link prediction focus on unsigned/unweighted social networks,and there are few researches on link prediction in WSSN,especially on edge weight prediction.This paper proposes an efficient weighted signed network embedding(WSNE)algorithm.For each directed edge,it learns the embedded representation of two endpoints and considers the edge symbol constraints to improve the performance of weight prediction.Compared with other related algorithms on three real WSSN datasets,the experimental results show that the WSNE can effectively predict the edge weight,and is superior to the related algorithms.