Imperial College London
Portrait Picture

DrSethFlaxman

Faculty of Natural Sciences, Department of Mathematics

Visiting Reader
 
 
 
//

Contact

 

s.flaxman

 
 
//

Location

 

522Huxley BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@article{Flaxman:2019:10.1214/19-AOAS1284,
author = {Flaxman, S and Chirico, M and Pereira, P and Loeffler, C},
doi = {10.1214/19-AOAS1284},
journal = {Annals of Applied Statistics},
pages = {2564--2585},
title = {Scalable high-resolution forecasting of sparse spatiotemporal events with kernel methods: a winning solution to the NIJ "Real-Time Crime Forecasting Challenge"},
url = {http://dx.doi.org/10.1214/19-AOAS1284},
volume = {13},
year = {2019}
}
Download

RIS format (EndNote, RefMan)

TY  - JOUR
AB  - We propose a generic spatiotemporal event forecasting method,which we developed for the National Institute of Justice’s (NIJ) RealTime Crime Forecasting Challenge (National Institute of Justice,2017). Our method is a spatiotemporal forecasting model combining scalable randomized Reproducing Kernel Hilbert Space (RKHS)methods for approximating Gaussian processes with autoregressivesmoothing kernels in a regularized supervised learning framework.While the smoothing kernels capture the two main approaches in current use in the field of crime forecasting, kernel density estimation(KDE) and self-exciting point process (SEPP) models, the RKHScomponent of the model can be understood as an approximation tothe popular log-Gaussian Cox Process model. For inference, we discretize the spatiotemporal point pattern and learn a log-intensityfunction using the Poisson likelihood and highly efficient gradientbased optimization methods. Model hyperparameters including quality of RKHS approximation, spatial and temporal kernel lengthscales,number of autoregressive lags, bandwidths for smoothing kernels, aswell as cell shape, size, and rotation, were learned using crossvalidation. Resulting predictions significantly exceeded baseline KDEestimates and SEPP models for sparse events.
AU  - Flaxman,S
AU  - Chirico,M
AU  - Pereira,P
AU  - Loeffler,C
DO  - 10.1214/19-AOAS1284
EP  - 2585
PY  - 2019///
SN  - 1932-6157
SP  - 2564
TI  - Scalable high-resolution forecasting of sparse spatiotemporal events with kernel methods: a winning solution to the NIJ "Real-Time Crime Forecasting Challenge"
T2  - Annals of Applied Statistics
UR  - http://dx.doi.org/10.1214/19-AOAS1284
UR  - http://hdl.handle.net/10044/1/71925
VL  - 13
ER  -
Download