Hendon Publishing - Article Archive Details

Print Article Rate Comment Reprint Information

Using GIS and statistical analysis to understand crime patterns

Written by Robert Cheetham

Although the modern police force has greater access than ever before to information about the incidence of crime through computer-aided dispatch (CAD) and records management systems (RMS), the sheer volume of data that must be sorted through on a daily basis can be staggering. Time and resource constraints make it difficult for law enforcement officials to properly analyze or disseminate the data in an organized and meaningful fashion. In the case of the Philadelphia Police Department, for example, the officers in the field are supported by a crime analysis unit, but the unit includes only three analysts to support a department of more than 6,600 sworn officers. To bridge gaps like this between ever-increasing levels of data and constrained resources, police departments have been seeking information management systems and visualization tools to help them sift through their available data and identify the most significant intelligence.

Geographic information systems, or GIS, play an increasingly important role in the integration and visualization of crime data for both law enforcement and public safety applications. However, preventing crime is a more complex task than simply mapping incidents or arrests and deploying resources accordingly—police departments must do more than just “follow the dots.” To be on the cutting edge of crime prevention, law enforcement agencies need geospatial data-mining tools that can automatically pull information out of incident databases and provide the means for detecting and stopping spikes in crime before they escalate further.

During the past several years, the National Institute of Justice (NIJ) has supported research into the development of advanced crime analysis techniques and tools for a variety of applications and purposes. Funding for similar research is also made available through other federal agencies. The City of Philadelphia and the Law Enforcement Support Agency for the City of Tacoma and Pierce County, Wash., are taking advantage of these initiatives by implementing Web-based GIS applications developed with grant funding from the National Science Foundation.

Crime Spike Detector

With the support of the Philadelphia Police Department and the Office of the U.S. Attorney, Azavea, a geospatial software design firm based in Philadelphia, developed a prototype software tool based on a paper that founder Robert Cheetham had presented at the 2000 Crime Mapping Research Conference. This geographic change detection system would sift through millions of records each day and identify statistically significant spikes in clusters of crime events. Called the Crime Spike Detector, it worked on a principle of comparing the density of crime from a recent period to a period of history in the past. In the Philadelphia police districts that tested the prototype, the system was able to discover crime hot spots that were split between districts and had otherwise gone unnoticed. The captains of these districts were so enthusiastic about the potential for the Crime Spike Detector that they began requesting the configuration of search patterns that would enable them to monitor less serious events such as domestic disturbances, vandalism and traffic accidents.

Before founding Azavea in 2000, Robert Cheetham was one of two founding analysts of what is now the Crime Analysis and Mapping Unit of the Philadelphia Police Department, so he had firsthand knowledge of the needs and concerns of urban law enforcement agencies. Since 2002, Azavea has built on this experience to develop crime analysis and mapping software tools for a range of organizations including the Philadelphia Police Department; the Law Enforcement Support Agency for the City of Tacoma and Pierce County, Wash.; the Bureau of Alcohol, Tobacco and Firearms; the Department of Justice Office of Juvenile Justice and Delinquency Prevention; and smaller agencies such as housing authority and university police departments. In 2007, Azavea leveraged this success to win a Small Business Innovation Research (SBIR) grant from the National Science Foundation to support the development of a nextgeneration Crime Spike Detector software tool that would ultimately be called HunchLab™.

Quantifying Police Hunches

HunchLab is a web-based software application that can apply advanced spatial statistics to identify pattern in crime data, notify law enforcement personnel with customized alerts using a geography-based subscription service, and assist them with visualizing those patterns and turning them into useful intelligence. Results are displayed as interactive maps, graphs and reports, including hot spot or “heat” maps that show the density of crime patterns over a specified geographic area and timeframe. Turning human theories about crime into quantifiable search parameters is a key aspect of HunchLab. Law enforcement officials must assemble data, hardwon personal experience, and local knowledge on a daily basis to make sense of these changing patterns and to formulate theories about criminal activity occurring at a particular place and time. These theories are called hunches, and they are developed and acted upon every day in police departments throughout the world, often with no formal mechanism to impartially verify or deny that they are valid. HunchLab makes it possible to test and quantify many types of hunches using the data collected during daily police department activities.

In HunchLab, a hunch is typically comprised of four separate pieces of information:

• Crime class and other filters such as narrative text and time of day

• Spatial extent of the area of interest, such as a police district or neighborhood

• Definition of the historical time period, such as last year at this time

• Definition of the recent time period, such as within the past week

HunchLab users are able to define the precise characteristics of each of these components to identify areas in which recent events demonstrate statistically significant changes from historic events. Once they are created, HunchLab can store thousands of these search patterns as part of an agency’s enterprise GIS. The data import and hunch processing services run on a schedule defined by system administrators. In Philadelphia, for example, these services are automatically run at night. When the services execute, records from the Philadelphia Police Department’s incident database are extracted, transformed and loaded (ETL) into the HunchLab database. The Hunch Processing service then sifts through the 25 million records in the Philadelphia Police Department’s incident database to identify statistically significant changes in geographic patterns of crime. Once a “crime spike” is detected, it triggers an automated e-mail alert sent to a geography-based subscription list of users.

HunchLab imports incident level data from multiple sources, such as a police department’s existing RMS, CAD and other applications. Users can view incidents from the different data sources either separately or in comparison to one another using a map, table or other analysis output. An interactive map enables users to zoom in, zoom out, or pan to different geographic areas to view selected incidents. In large urban areas, where a number of like incidents may occur in close proximity, the incidents are clustered together so the map remains clean and easy to read. If multiple types of incidents are being viewed, the clusters will appear as pie charts showing the relative distribution of each type of incident within the cluster location. By clicking on one of these incidents or clusters of incidents, a user can access more detailed information about the criminal activity that occurred there. Incident data can also be displayed as a heat map to graphically visualize trends and patterns in crime.

HunchLab’s subscription and alerting system manages the database of hunches, the requested alert timing and parameters, and the people who are to be notified of changes in related crime activity. Users can subscribe to any hunch in the database, regardless of creator, and receive e-mail alerts when activity changes. Hunches are also searchable by multiple criteria.

A Proactive Approach to Policing

The overarching goal of the HunchLab application is to enable a more proactive deployment of available staff time and resources in order to manage recurring crime incidents more effectively. Azavea’s ongoing research for developing, enhancing and improving the application has involved extensive interviews and discussions with key members of the law enforcement community including the Philadelphia Police Department, the Law Enforcement Support Agency, the Department of Homeland Security, the International Association of Chiefs of Police, the ESRI Law Enforcement Team, and teachers and students from local universities.

“Both the original Crime Spike Detector and the new HunchLab software have proven to be stateof- the-art analytical and pattern detection tools for the Philadelphia Police Department,” said Michael Urciuoli, GIS Specialist with the Crime Analysis and Mapping Unit. “Philadelphia police personnel love the combination of a hands-off tool that can comb through millions of records nightly while also providing easy-to-use Web-based tools for analyzing and visualizing the clusters that are found.”

The Philadelphia Police Department reports that thanks to HunchLab they have managed to streamline their processes, enabling officers to communicate more effectively and efficiently on crime activity occurring in the city. Because HunchLab is web-based, they are able to create hunches themselves and share them with other members of their district or special units. Detectives, officers and supervisors can “subscribe” to these hunches and help monitor them and act upon them. The Philadelphia Housing Department also leverages the application to provide daily e-mail digests of crime incidents that occur on Housing Department properties.

The Philadelphia Police Department’s HunchLab application is built to leverage ESRI ArcGIS Server technology and can also be integrated with ArcGIS Online. Other deployment options include Google Maps and Microsoft Bing Maps, depending on an organization’s preference or available technology. For organizations that prefer open source technology, HunchLab can also be integrated with the GeoServer Web mapping system. The HunchLab application is compatible with the commodity hardware and software used by many police departments.

The HunchLab implementation for the Law Enforcement Support Agency for the City of Tacoma and Pierce County, Wash., is part of a larger system of crime mapping and analysis tools known as the Crime Early Warning System, or CEWS (pronounced C-Wiz). The CEWS project was funded by an Edward Byrne Memorial Discretionary Grant from the United States Department of Justice. Unlike the Philadelphia implementation of HunchLab, where a single law enforcement agency has jurisdiction over both the City and County of Philadelphia, the CEWS application uses data from multiple, separate jurisdictions and was designed to potentially integrate with additional jurisdictions in the future.

Final installation and beta testing of the CEWS application was completed in early 2010 and has already generated a dramatic early success story involving a small island community in Puget Sound. The island is under the jurisdiction of Pierce County and is accessible only by boat or ferry. It has a year-round population of approximately 1,100 residents that swells to more than 4,000 residents during peak summer months. A sudden rise in burglaries on the island was under investigation while the initial capabilities of CEWS were being demonstrated. Pierce County Sheriff’s Sergeant Mike Blair was very impressed with the geographic representation of the island’s burglary spike, but even more so with the Time-of-Day/Day-of-Week Heat Map graph extracted from the HunchLab side of the application. By studying this graph and the related incident data, Sergeant Blair immediately realized that the majority of incidents were occurring in the hours between mid-morning and noon, so his original plan to send a patrol car to the island on swing shift was not likely to catch the burglar. He shared the CEWS data with a seasoned deputy from the Pierce County Sheriff’s Department, who in turn drew on his instincts and investigative skills to change the deployment pattern. The serial burglar responsible for the spike was apprehended within 24 hours.

Looking Ahead

In 2010, Azavea received Phase IIB supplemental grant funding from the National Science Foundation that will facilitate further enhancements to HunchLab, including new risk forecasting tools that will support the National Institute of Justice’s Information-Led Policing initiative. Research has proven there is a near-repeat phenomenon in certain crime patterns for repeat victimization or retaliatory actions in or around the area where the initial or parent event took place. Azavea is currently working with Dr. Jerry Ratcliffe at Temple University’s School of Criminal Justice to integrate his work with the near-repeat pattern and seasonality forecasting into the next iteration of HunchLab.

Robert Cheetham is the founder and president of Azavea. Previously, he was a software developer and GIS analyst for the Philadelphia Police Department, the University of Pennsylvania and the City of Philadelphia. He can be reached at cheetham@ azavea.com.

Mary L. Johnson is the Technical Writer for Azavea. She has over 25 years of experience and has previously worked in the civil engineering and petroleum industries. She can be reached at mjohnson@azavea.com. Photos courtesy of Azavea.

Published in Public Safety IT, Jan/Feb 2011

Rating : 10.0


Comments

Comment on This Article

No Comments


Related Companies

Azavea
 

Related Products

CAD (Computer Aided Dispatch)Records ManagementRMS (Record Management System)
 

Article Images

Click to enlarge images.

Events and Tradeshows: LAOPFMTRPSIT
Latest News: LAOPFMTRPSIT
 
Close ...