【摘要】 复杂网络节点的重要性度量与时间属性相关，经典静态网络模型弱化对节点交互时间属性的有效表征.将深度学习模型迁移到动态图数据上进行端到端系统建模，提出基于图卷积融合计算的时效网络节点重要性综合评估模型.通过超邻接矩阵集结时效网络结构特征的动态演化过程，利用图卷积神经网络框架融合计算节点邻域特征，分析节点时序演化重要性顺序结构，实现节点重要性综合排序.仿真实验结果表明，与基线方法相比，所提方法得到的Kendall’sτ值在所选网络数据集上均表现优良，体现出基于图卷积融合计算的时效网络节点重要性综合评估方法的有效性和优越性.更多还原

【Abstract】 The importance measure of nodes in complex networks is correlated with the time attribute. The classical static network model weakens the effective representation of the time attribute of node interaction. A node importance evaluation model for temporal networks based on the graph convolution union computing was proposed.The model migrated the deep learning to dynamic graph data for end-to-end system modeling. Dynamic evolution process of the temporal network structure was assembled by the supra-adjacency matrix. The graph convolutional neural network framework was used to calculate the fusion characteristics of the neighborhood nodes. The node importance order structure over time was analyzed. A comprehensive ranking of node importance was achieved. The simulation experimental results showed that compared with the existing method, the Kendall’s tau values obtained by the proposed method performed well on all the selected network datasets, reflecting the effectiveness and superiority of the proposed method.更多还原