Deep learning for real-time crime forecasting and its ternarization

DOI10.1007/S11401-019-0168-YMaRDI QIDQ2286236zbMATH OpenOpenAlexWikidataFDO

Authors Bao Wang, Penghang Yin, P. Jeffrey Brantingham, Jack Xin, Andrea L. Bertozzi, Stanley Osher

Publication date 10 January 2020

Published in Chinese Annals of Mathematics. Series B (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1711.08833

zbMATH Keywords

real-time forecastingcrime representationspatial-temporal deep learningternarization

Mathematics Subject Classification ID

Artificial neural networks and deep learning (68T07) Mathematical modeling or simulation for problems pertaining to game theory, economics, and finance (91-10) Mathematical sociology (including anthropology) (91D99) Computational methods for problems pertaining to game theory, economics, and finance (91-08)

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, we first present a proper representation of crime data. We 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, we 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 et al, Arxiv 1707.03340].

Recommendations

• Scalable high-resolution forecasting of sparse spatiotemporal events with kernel methods: a winning solution to the NIJ ``Real-time crime forecasting challenge
• Predicting citywide crowd flows using deep spatio-temporal residual networks
• Personalized crime location prediction
• Spatio-temporal low count processes with application to violent crime events
• Sequential data assimilation for 1D self-exciting processes with application to urban crime data

Cites work

• A STATISTICAL MODEL OF CRIMINAL BEHAVIOR
• Gang rivalry dynamics via coupled point process networks
• Nonlinear Patterns in Urban Crime: Hotspots, Bifurcations, and Suppression
• Self-exciting point process modeling of crime

Describes a project that uses

Uses Software

• LaplacianSmoothing-GradientDescent
• XNOR-Net
• BinaryNet
• BinaryConnect
• Adam
• cuDNN
• Theano

This page was built for publication: Deep learning for real-time crime forecasting and its ternarization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2286236)