Spatio Temporal Modelling of Repeat Crimes and Hotspots Prediction Using Recurrent Deep Neural Networks

Recent research on criminology has given rise to an increased interest on crime analysis, especially crime hotspot analysis. In this aspect, the understanding of repeat crimes become more essential as there is always an increased risk of a location being targeted again after an initial incident. In most of the traditional crime analysis approaches, both time and location are considered as separate entities. This work proposes to model repeat crimes by using space and time as interdependent entities. The focus of this paper is to provide a comprehensive spatio-temporal analysis of repeat crimes and prediction of crime hotspots using Machine learning & Deep learning algorithms. The hotspot prediction model was built using different state-of-the-art machine learning algorithms. However, Recurrent Deep neural network model provided the highest accuracy of $96.5 \%$ as compared to the other algorithms.