U.S. flag

An official website of the United States government, Department of Justice.

Dot gov

The .gov means it’s official.
Federal government websites always use a .gov or .mil domain. Before sharing sensitive information online, make sure you’re on a .gov or .mil site by inspecting your browser’s address (or “location”) bar.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Point process modeling with spatiotemporal covariates for predicting crime

Award Information

Award #
2016-R2-CX-0021
Location
Congressional District
Status
Closed
Funding First Awarded
2016
Total funding (to date)
$99,096

Description of original award (Fiscal Year 2016, $49,548)

As submitted by the proposer: Statement of the Problem. Criminologists and law enforcement agencies want to better predict crime hotspots and understand the factors that cause them. Prior research suggests that past crime hotspots, spatial features (e.g. bars or public transit stops), leading indicators (e.g. public drunkenness or 911 calls), and weather are all predictive of future crime, but no proposed predictive policing model uses all of these factors, so their effects are difficult to compare. Model usefulness is limited because most methods do not predict crime rates along with hotspot locations. Methods to evaluate model prediction accuracy are also limited, leaving practitioners uncertain which methods are most useful for predicting crime. A unified model and evaluation method could suggest proactive police tactics, answer significant criminological questions, and inform public policy.
Research Design and Methods. We will extend self-exciting point process models to combine features of hotspot, leading indicators, and Risk Terrain Modeling, producing a single unified model which can predict crime using recent crime data, spatial covariates, and temporal covariates, such as weather or season. Unlike in previous predictive methods, tuning parameters (such as spatial bandwidths) will be determined empirically from past local crime data instead of estimated by the operator. The model will be implemented as open-source software and used to develop statistical methods to evaluate model fit and predictive accuracy, and to test which covariates or leading indicators significantly predict future crime.
Partnerships. We are collaborating with the Pittsburgh Bureau of Police, using their expertise to guide our analysis and using local crime data to test our methods.
Analysis. Our model will be used to analyze publicly available crime data from Pittsburgh, PA and other major cities to determine which factors best predict violent crime. By comparing results between major cities, we hope to understand which relationships are universal and which vary based on location, policing tactics, demographics, and other city-to-city variations. We will also explore extending the model to be hierarchical, combining information from multiple cities to better understand variations in crime.
Products, Reports, and Data Archiving. Findings will be published in a thesis, peer-reviewed articles, and presentations at professional conferences. Open-source software implementing our analysis will be made available online.

Note: This project contains a research and/or development component, as defined in applicable law.
ca/ncf

As submitted by the proposer: Criminologists and law enforcement agencies want to better predict crime hotspots and understand the factors that cause them. Prior research suggests that past crime hotspots, spatial features (e.g. bars or public transit stops), leading indicators (e.g. public drunkenness or 911 calls), and weather are all predictive of future crime, but no proposed predictive policing model uses all of these factors, so their effects are difficult to compare.

Model usefulness is limited because most methods do not predict crime rates along with hotspot locations. Methods to evaluate model prediction accuracy are also limited, leaving practitioners uncertain which methods are most useful for predicting crime.

A unified model and evaluation method could suggest proactive police tactics, answer significant criminological questions, and inform public policy. Research Design and Methods. The researcher will extend self-exciting point process models to combine features of hotspot, leading indicators, and Risk Terrain Modeling, producing a single unified model which can predict crime using recent crime data, spatial covariates, and temporal covariates, such as weather or season.

Unlike in previous predictive methods, tuning parameters (such as spatial bandwidths) will be determined empirically from past local crime data instead of estimated by the operator. The model will be implemented as open-source software and used to develop statistical methods to evaluate model fit and predictive accuracy, and to test which covariates or leading indicators significantly predict future crime.

The researcher is collaborating with the Pittsburgh Bureau of Police, using their expertise to guide our analysis and using local crime data to test our methods.

The model will be used to analyze publicly available crime data from Pittsburgh, PA and other major cities to determine which factors best predict violent crime. By comparing results between major cities, the researcher hopes to understand which relationships are universal and which vary based on location, policing tactics, demographics, and other city-to-city variations. The researcher will also explore extending the model to be hierarchical, combining information from multiple cities to better understand variations in crime. Products, Reports, and Data Archiving. Findings will be published in a thesis, peer-reviewed articles, and presentations at professional conferences. Open-source software implementing our analysis will be made available online.

This project contains a research and/or development component, as defined in applicable law. This project contains a research and/or development component, as defined in the applicable law, “and complies with Part 200 Uniform Requirements – 2 CFR 200.210(a)(14). nca/ncf

Date Created: July 17, 2016