STGSN — A Spatial–Temporal Graph Neural Network framework for time-evolving social networks
Shengjie Min,Zhan Gao,Jing Peng,Liang Wang,Ke Qin,Bo Fang
DOI: https://doi.org/10.1016/j.knosys.2021.106746
2021-02-01
Abstract:<p>Social Network Analysis (<strong>SNA</strong>) has been a popular field of research since the early 1990s. Law enforcement agencies have been utilizing it as a tool for intelligence gathering and criminal investigation for decades. However, the graph nature of social networks makes it highly restricted to intelligence analysis tasks, such as role prediction (node classification), social relation inference (link prediction), and criminal group discovery (community detection), etc. In the past few years, many studies have focused on Graph Neural Network (<strong>GNN</strong>), which utilizes deep learning methods to solve graph-related problems. However, we have rarely seen GNNs tackle time-evolving social network problems, especially in the criminology field. The existing studies have commonly over-looked the temporal-evolution characteristics of social networks. In this paper, we propose a graph neural network framework, namely <strong>S</strong>patial-<strong>T</strong>emporal <strong>G</strong>raph <strong>S</strong>ocial <strong>N</strong>etwork (<strong>STGSN</strong>), which models social networks from both spatial and temporal perspectives. Using a novel approach, we leverage the temporal attention mechanism to capture social networks' temporal features. We design a method analyzing temporal attention distribution to improve the interpretation ability of our method. In the end, we conduct extensive experiments on six public datasets to prove our methods' effectiveness.</p>
computer science, artificial intelligence