Spatio-Temporal Deep Fusion Graph Convolutional Networks for Crime Prediction

Abstract

Effective crime prediction plays a key role in sustaining the stability of society. In recent years, researchers have proposed a number of prediction methods that extract spatial and temporal features separately and fuse afterward. However, the strict distinction between spatial feature extraction and temporal feature extraction can result in the loss of useful information. To this end, we propose a spatio-temporal deep fusion graph convolution network (STDGCN), which embodies the intra-region spatio-temporal features and the inter-region spatio-temporal associations on a single graph. STDGCN performs the convolution without distinguishing between space and time to simultaneously extract spatio-temporal features. Our evaluations of two real-world datasets demonstrate the effectiveness of STDGCN.