时序超网络上重要节点挖掘方法研究

在如何识别时序超网络上的重要节点方面取得了一定的进展。定义了该类网络上度量节点重要性程度的8个中心性方法及随机移除节点的基线方法，分别侧重于网络不同的拓扑结构性质和时间特征，从多个角度综合考虑了该类网络上节点的重要性。同时，构建了时序超网络上的SI传播模型，基于该模型提出了新的评估方法来衡量所提出的中心性方法的有效性。研究表明，在时序超网络上，基于最快到达路径的介数中心性方法是评价该类网络上节点重要性的良好指标。此外，基于时间分辨率的度和超度中心性方法通过寻找网络的最佳时间分辨率，可以进一步优化普通的度和超度中心性方法，弥补了普通方法不能有效考虑网络时间信息的缺点，且在多个真实网络上表现出与介数中心性方法相当的性能。

Mining methods of important nodes in temporal hyper-networks

Some progress has been made on how to identify important nodes on the temporal hyper-network. Eight centrality methods for measuring the importance of nodes on this type of network and a baseline method for randomly removing nodes were defined, focusing on the different topological properties and time characteristics of the network, and comprehensively considering the importance of nodes on this type of network from multiple perspectives. At the same time, a SI spreading model on temporal hyper-network was constructed. Based on this model, a new evaluation metric was put forward to measure the effectiveness of the proposed centrality metrics. The main results are as follows. Betweenness metric which is based on the fastest arrival path performs better than the other baselines in identifying important nodes in temporal hyper-networks. In addition, the degree and hyper-degree centrality which consider time resolution are superior to the degree and hyper-degree based on the static network topology, which makes up for the shortcomings that the static methods that can not effectively consider the network time information. In addition, they show comparable performance with the betweenness in multiple real networks.