Multi-Dimensional HITS Based on Random Walks for Multilayer Temporal Networks

Numerous centrality measures have been established to identify the important nodes in static networks, among them, HITS centrality is widely used as a ranking method. In this paper, we extend the classical HITS centrality to rank nodes in multilayer temporal networks with directed edges. First, we use a sixth-order tensor to represent multilayer temporal network and then introduce random walks in the established sixth-order tensor by constructing six transition probability tensors. Second, we establish tensor equations based on these constructed tensors to obtain six centrality vectors: two for the nodes, two for the layers and two for the time stamps. Besides, we prove the existence of the proposed centrality measure under some conditions. Finally, we experimentally show the effectiveness of the proposed centrality on an synthetic network and a real-world network.