Self-Exciting Point Processes with Spatial Covariates: Modelling the Dynamics of Crime

Reinhart, Alex; Greenhouse, Joel B.

doi:10.1111/rssc.12277

Cited by 43 publications

(21 citation statements)

References 62 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…'predictive policing' (Perry et al 2013). Reinhart and Greenhouse (2018) classify the methods most often used as belonging to one of three classes: (i) hot spot maps, (ii) near repeat analysis, (iii) regression-based methods.…”

Section: The Practice Of Predicting Crimementioning

confidence: 99%

Mapping the Risk Terrain for Crime Using Machine Learning

Wheeler

Steenbeek

2020

J Quant Criminol

View full text Add to dashboard Cite

Objectives: We illustrate how a machine learning algorithm, Random Forests, can provide accurate long-term predictions of crime at micro places relative to other popular techniques. We also show how recent advances in model summaries can help to open the 'black box' of Random Forests, considerably improving their interpretability. Methods: We generate long-term crime forecasts for robberies in Dallas at 200 by 200 feet grid cells that allow spatially varying associations of crime generators and demographic factors across the study area. We then show how using interpretable model summaries facilitate understanding the model's inner workings. Results: We find that Random Forests greatly outperform Risk Terrain Models and Kernel Density Estimation in terms of forecasting future crimes using different measures of predictive accuracy, but only slightly outperform using prior counts of crime. We find different factors that predict crime are highly non-linear and vary over space. Conclusions: We show how using black-box machine learning models can provide accurate micro placed based crime predictions, but still be interpreted in a manner that fosters understanding of why a place is predicted to be risky.

show abstract

Section: The Practice Of Predicting Crimementioning

confidence: 99%

Mapping the Risk Terrain for Crime Using Machine Learning

Wheeler

Steenbeek

2020

J Quant Criminol

View full text Add to dashboard Cite

show abstract

“…Despite the large sample sizes, tests of crime reductions tended to only show a statistically significant crime reduction effect by a small margin. One potential way to account for this in future research is to estimate the decay effect directly, such as by decomposing the offspring effect in a Hawkes model into two components, crimes with and without arrests (Achab et al, 2018;Mohler, Carter, & Raje, 2018;Reinhart & Greenhouse, 2018).…”

Section: Discussionmentioning

confidence: 99%

Breaking the chain: How arrests reduce the probability of near repeat crimes

Wheeler¹,

Riddell²,

Haberman³

2019

Preprint

View full text Add to dashboard Cite

Objectives: Near repeat patterns have been identified for a host of different crimes, but effective strategies to reduce near repeats have had more variable results. This study identifies near repeat crime patterns in Dallas, Texas and examines the effects of an arrest on reducing the probability of future crime. Method: Using open source crime data from the Dallas Police Department from July 2014 through June 2018, we identified near repeat patterns for shootings, interpersonal robberies, residential burglaries, and thefts from motor vehicles. Logistic regression models were used to test the effect of an arrest on reducing near repeat crimes; controls for geographic, demographic, and temporal factors were included in each model. Results: Near repeat calculations suggest violent crime clustered closely in time and space, with property crime dispersed over larger spatial and temporal dimensions. Across all four crime types, findings suggest arrests resulted in 20% to 40% reductions in a near repeat follow up crime.Conclusions: In line with past research on shootings, arrests reduced the likelihood of subsequent crimes. This suggests policing strategies to increase arrests may be a fruitful way to reduce near repeat crime patterns.Code to replicate the analysis can be found at:https://www.dropbox.com/sh/72q0i7wm60ndqve/AABgFB8zwGHG6xjExvaD01EGa?dl=0.

show abstract

“…Comparing with the results of the analysis of the burglary crimes in Los Angeles during the period of Mohler et al (2011), the clustering effect in the robbery violence data seems much lower. In Reinhart and Greenhouse (2018), the proportion of clustering events in all the burglary crimes in Pittsburgh during 2011-2016 amounts to 47%. The reason might be that the same burglar watches and visits several neighbouring houses within a short timespan, whereas a robber always escapes from the crime spot quickly to avoid being caught.…”

Section: Model Fittingmentioning

confidence: 99%

“…By adopting the formulation of the Hawkes process, the model of Mohler and co-workers incorporates the time varying hotspots and near repeats with the assumption that every crime induces a locally higher risk of crime which decays in space and time. Reinhart and Greenhouse (2018) considered a background with simple spatial covariates. Since parametric models are difficult to construct for data where empirical studies are insufficient, non-parametric and semiparametric estimation methods for the Hawkes model have been developed.…”

Section: Introductionmentioning

confidence: 99%

A Semiparametric Spatiotemporal Hawkes-Type Point Process Model with Periodic Background for Crime Data

Zhuang

Mateu

2018

Journal of the Royal Statistical Society Series A: Statistics in Society

View full text Add to dashboard Cite

Summary Past studies have shown that crime events are often clustered. This study proposes a spatiotemporal Hawkes‐type point process model, which includes a background component with daily and weekly periodization, and a clustering component that is triggered by previous events. We generalize the non‐parametric stochastic reconstruction method so that we can estimate each component in the background rate and the triggering response that appears in the model conditional intensity: the background rate includes a daily and a weekly periodicity, a separable spatial component and a long‐term background trend. Two relaxation coefficients are introduced to stabilize and secure the estimation process. This model is used to describe the occurrences of violence or robbery cases in Castellón, Spain, during 2 years. The results show that robbery crime is highly influenced by daily life rhythms, revealed by its daily and weekly periodicity, and that about 3% of such crimes can be explained by clustering. Further diagnostic analysis shows that the model could be improved by considering the following ingredients: the daily occurrence patterns are different between weekends and working days; in the city centre, robbery activity shows different temporal patterns, in both weekly periodicity and long‐term trend, from other suburb areas.

show abstract

Self-Exciting Point Processes with Spatial Covariates: Modelling the Dynamics of Crime

Cited by 43 publications

References 62 publications

Mapping the Risk Terrain for Crime Using Machine Learning

Mapping the Risk Terrain for Crime Using Machine Learning

Breaking the chain: How arrests reduce the probability of near repeat crimes

A Semiparametric Spatiotemporal Hawkes-Type Point Process Model with Periodic Background for Crime Data

Contact Info

Product

Resources

About