Spatiotemporal Prediction of Theft Risk with Deep Inception-Residual Networks

Spatiotemporal prediction of crime is crucial for public safety and smart cities operation. As incidents are distributed sparsely across space time, existing deep-learning methods constrained by coarse spatial scale offer only limited values in density. This paper proposes the use deep inception-residual networks (DIRNet) to conduct fine-grained, theft-related based on non-emergency service request data (311 events). Specifically, it outlines employment inception units comprising asymmetrical convolution layers draw low-level spatiotemporal dependencies hidden events complaint records 311 dataset. Afterward, this details how residual can be applied capture high-level features from final prediction. The effectiveness proposed DIRNet evaluated New York City 2010 2015. results confirm that obtains an average F1 71%, which better than other models.