CrimeSTC: A Deep Spatial-Temporal-Categorical Network for Citywide Crime Prediction

Abstract

Crime is one of the most complex social problems around the world, posing a major threat to human life and property. Predicting crime incidents in advance can be a great help in fighting against crime and has drawn continuous attention from both academic and industrial communities. Although a plethora of methods have been proposed over the past decade, most of the algorithms either perform prediction by leveraging linear or other oversimplified models or fail to fully explore the dynamic patterns in the crime data. In this paper, we propose a novel deep learning based crime prediction framework called CrimeSTC to jointly learn the intricate spatial-temporal-categorical correlations hidden inside the crime and big urban data. Specifically, our framework consists of four parts: dynamic module (handling the data that change every day via local CNN and GRU), static module (handling the data that remain the same over time via fully connected layers), categorical module (capturing the categorical dependency via graph convolutional network) and joint training module (concatenating dynamic and static representations to forecast crime numbers). Extensive experiments on real world datasets validate the effectiveness of our framework.