Abstract
Recently, temporal graphs have substituted dynamic graphs as many real-world problems evolve in continuous time rather than in discrete time, and besides time almost all problems are designed in a heterogeneous format rather than a homogeneous one. However, most existing graph representations do not consider time in their components. To this end, in this paper, we present a new heterogeneous graph representation including time in every single component of the graph, i.e., nodes and edges. We also introduce four time-dependent queries to address machine learning or deep learning problems. Our findings reveal that considering the size of the enormous graphs, our time-dependent queries execute efficiently. In order to show the expressive power of time in graph representation, we construct a graph for a new social media platform (Steemit), and address a DL prediction task using graph neural networks (GNNs). Predicting the payout for a newly published post is one of the most fascinating classification problems in the Steemit setting, and we address this problem with two approaches followed by GNN models.
Abstract (translated)
URL
https://arxiv.org/abs/2209.03144
