Deep Learning for Real-Time Crime Forecasting and Its Ternarization

Abstract

Real-time crime forecasting is important. However, accurate prediction of when and where the next crime will happen is difficult. No known physical model provides a reasonable approximation to such a complex system. Historical crime data are sparse in both space and time and the signal of interests is weak. In this work, the authors first present a proper representation of crime data. The authors then adapt the spatial temporal residual network on the well represented data to predict the distribution of crime in Los Angeles at the scale of hours in neighborhood-sized parcels. These experiments as well as comparisons with several existing approaches to prediction demonstrate the superiority of the proposed model in terms of accuracy. Finally, the authors present a ternarization technique to address the resource consumption issue for its deployment in real world. This work is an extension of our short conference proceeding paper [Wang, B., Zhang, D., Zhang, D. H., et al., Deep learning for real time Crime forecasting, 2017, arXiv: 1707.03340].

Keywords

Crime representation / Spatial-temporal deep learning / Real-time forecasting / Ternarization

References

[1] Chen P, Yuan H, Shu X. Forecasting crime using the arima model. Proceeding of the 5th IEEE International Conference on Fuzzy Systems and Knowledge Discovery, 2008, 5: 627-630

[2] Chen X, Cho Y, Jang S. Crime prediction using twitter sentiment and weather. Systems and Information Engineering Design Symposium, 2015 63-68

[3] Chetlur S, Woolley C, Vandermersch P cuDNN: Efficient primitives for deep learning, 2014

[4] Choillet F. Keras: Keep learning for humans, 2015

[5] Courbariaux M, Bengio Y, David J. Binaryconnect: Training deep neural networks with binary weights during propagations. Advances in Neural Information Processing Systems, 2015, 28: 3123C3131

[6] Courbariaux M, Hubara I, Soudry D Binarized neural networks: Training neural networks with weights and activations constrained to +1 or -1, 2016

[7] Dieleman S, Schlter J, Raffel C Lasagne: First release, 2015

[8] Gerber M. Predicting crime using twitter and kernel density estimation. Decision Support System, 2014, 61: 115-125

[9] He K M, Zhang X Y, Ren S Q, Sun J. Deep residual learning for image recognition. CVPR, 2016 770-778

[10] Hochreiter S, Schmidhuber J. Long short-term memory. Neural Comput, 1997, 9: 1735-1780

[11] Holden Daniel, Komura Taku, Saito Jun. Phase-functioned neural networks for character control. ACM Transactions on Graphics, 2017, 36(4): 1-13

[12] Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift, 2015

[13] Jain A, Zamir A R, Savarese S, Saxena A. Structural-rnn: Deep learning on spatio-temporal graphs. CVPR, 2016

[14] Kang Hyeon-Woo, Kang Hang-Bong. Prediction of crime occurrence from multi-modal data using deep learning. PLOS ONE, 2017, 12(4): e0176244

[15] Kingma D P, Ba J. Adam: A method for stochastic optimization. ICLR, 2015

[16] LeCun Y, Bengion Y, Hinton C. Deep learning. Nature, 2015, 521: 436-444

[17] Li F, Zhang B, Liu B. Ternary weight networks. NIPS Workshop, 2016

[18] Li H, De S, Xu Z Training quantized nets: A deeper understanding, 2017

[19] Li Y, Zemel R, Brockschmidt M, Tarlow D. Gated graph sequence neural network. ICLR, 2016

[20] Mohler G O, Short M B, Brantingham P J Self-exciting point process modeling of crime. J. Amer. Statist. Assoc, 2011, 106(493): 100-108

[21] Mohler, G. O., Short, M. B. and Brantingham, P. J., The concentration dynamics tradeoff in crime hot spotting, Unraveling the Crime-Place Connection: New Directions in Theory and Policy., 22, 2017, 21 pages.

[22] Osher S, Wang B, Yin P Laplacian smoothing gradient descent, 2018

[23] Rastegari M, Ordonez V, Redmon J, Farhadi A. Xnor-net: Imagenet classification using binary convolutional neural networks. EGGV, 2016

[24] Short M B, Mohler G O, Brantingham P J, Tita G E. Gang rivalry dynamics via coupled point process network. Discrete Contin. Dyn. Syst. Ser. B, 2014, 19(5): 1459-1477

[25] Short M B, Bertozzi A L, Brantingham P J. Nonlinear patterns in urban crime: Hotspots, bifurcations, and suppression. SIAM J. Appl. Dyn. Syst., 2010, 9(2): 462-483

[26] Short M B, Brantingham P J, Bertozzi A L, Tita G E. Dissipation and displacement of hotspots in reaction-diffusion models of crime. Proc. Nat. Acad. Sci, 2010, 107(9): 3961-3965

[27] Short M B, D'Orsogna M R, Pasour V B Persistent heat signature for pose-oblivious matching of incomplete models. M3AS: Mathematical Models and Methods in Applied Sciences, 2008, 18: 1249-1267

[28] Inverse Problems, 2011, 27 11

[29] The Theano Development Team Theano: A python framework for fas computation of mathematical expressions, 2015

[30] Wang B, Zhang D, Zhang D H Deep Learning for Real Time Crime Forecasting, 2017

[31] Wang X, Gerber M S, Brown D E. Automatic crime prediction using events extracted from twitter posts. International Conference on Social Computing, Behavioral-Cultural Modeling, and Prediction, 2012 231-238

[32] Yin P, Zhang S, Xin J, Qi Y. Quantization and Training of Low Bit-Width Convolutional Neural Networks for Object Detection, 2016

[33] Zhang J B, Zheng Y, Qi D R. Deep spatio-temporal residual networks for citywide crowd flows prediction. AAAI, 2017

[34] Zhou A J, Yao A B, Guo Y W Incremental network quantization: Towards lossless cnns with low-precision weights. ICLR, 2017

[35] Zhu C Z, Han S, Miao H Z, Dally W J. Trained ternary quantization. ICLR, 2017