Most police officers know that a large percentage of crime occurs in certain geographical areas and new technology is making it possible to predict crime trends and patterns and better allocate resources.
Motorola’s Intelligence Led Public Safety Survey found that 70 percent of police departments expect to implement predictive policing within 2–5 years and they anticipate a significant impact on reducing crime.
D. J. Seals, Industry Expert, “Predictive policing is not just historical, trending, or hotspotting. It uses intelligent analytics and algorithms in order to take multiple data sets into consideration. It’s not just looking at patterns but applying the analytics to current data to turn it into something consumable for officers on the road. Predictive policing makes sense for every officer because it is effective and applicable in real life.”
CommandCentral Predictive’s set of data looks at when, where, how often, the geographical area, whether crime moves around to other areas and the seasonality of crime. Every law enforcement officer knows that certain areas have more of a propensity for crime. What they may not know is whether or not there is a pattern, whether it moves around or is seasonal. Crime always spikes at Christmas, for example, but prior to this solution, we didn’t know the historical pattern, such as when it started in prior years, what type of crime, or where.
The algorithmic set tracks the functions all at once automatically. You can’t accurately make predictions based on one set of data so the solution uses all the functions to be considered in the algorithmic set, such as when, where, the movement of the crimes, and all other available data. The solution also breaks down the predictions by shift and ranked by their importance for that shift. Officers read the narratives of these ranked predictions of the most recent crimes, giving them tactically effective information at their fingertips empowering them to know what to do or look for in each of the prediction areas.
Seals added, “Any good officer can tell you where the majority of calls are, the difficulty comes in knowing when and where that next crime may occur. Officers are answering calls most of the time but when they aren’t, they need to be looking forward with predictive intelligence data, to determine where crime is more likely to occur and focus their energies there.”
The Department gets daily predictions on where crimes are most likely to occur on any given day, and patrol shift, constantly pulling in information from the RMS and data from other data bases. Departments use the analytic part of the tool, CommandCentral historical data and can then go to the predictive side and ask for predictions for such things as burglaries.
Seals stated, “It is futile to give an all-day prediction and part of the algorithmic set breaks down the data by the department’s shifts. The predictive data is for the shift being briefed. It is not effective to give predictions for the whole day; they must be broken down by shift so it is more applicable to officers on the road.”
Seals explained Targeted Area Predictions (TAP), “While other solutions offer just a geographical box, CommandCentral Predictive provides the most recent historical data that has occurred there. Then the analytics can provide type of crime information, vehicle, and suspect information. If we stopped at a box, we would just be hoping to be in the right place at the right time to stop a crime, but this really isn’t a win because the suspect will just return when no officer is there, or go to another area to commit that crime.
“We give you an intelligence narrative and parameters to check out, such as having a car or suspects to watch for and check out. The goal is stopping crime, not just pushing it around. I don’t see how predictive policing can work without an analytic and predictive function working together.”
CommandCentral Predictive and CommandCentral Analytics combine to create the complete predictive view. Predictive takes the data and adds the algorithm. There may be numerous boxes and officers need to know how to pick the ones in which to focus patrol. Boxes are tiered with rankings from mathematical predictions, the highest ranked being ‘1’ and officers focus on the highest ranking boxes but they can also be observant when they are entering boxes with ‘2’ or ‘3’ rankings.
Motorola put a lot of thought into design. The solution is completely mobile and available on any computer, mobile computer, smartphone, tablet, or other mobile device. Beat officers are able to get effective predictions and it is easily understandable and consumable.
Former San Diego Police Officer and Sheriff’s Deputy Kurt Smith, Director of Business Development, with TriTech Software, stated that his passion is connecting information with ideas, that data driven is not enough, and that information needs to be put into perspective. “TriTech’s customers understand they connect knowledge with action. The critical element is connecting knowledge to action to achieve real outcomes.”
Smith reported, “Any agency wanting to make a better use of information can come to us and we show and share what the command professionals can do to make changes, since improving information and analytics can identify the best opportunities for change or adjustment. There is no cookie cutter approach. The agency defines their needs, and TriTech puts the data to use and shows them how to evolve.”
TriTech accesses data and brings offender information to CrimeView to create a model that ranks offenders however an agency requests, with parameters such as prior police contacts, age, where they live, and how they have evolved as a criminal for certain types of crimes to develop crime predictions.
For instance, someone with a traffic contact in an area with a high burglary rate who has had drug-related contacts in the past might get flagged for a closer look. This is all tailored to the individual agency’s needs and requests. CrimeView provides a flexible and thorough way of putting information together with community information illuminating hard to determine patterns.
Smith stated, “The difference is that every department used to know the exact numbers of how much crime has occurred, but they now understand the contact and pattern and what has changed for the offender and victims. The Pareto Principle that 20 percent of the people commit 80 percent of the crimes is also true for geographical areas. Serious and prolific offenders are attracted to problem areas.”
CrimeView data can be shared if two or more agencies have CrimeView. Other agencies combine resources for regional Dashboards to catch crooks bouncing from one small jurisdiction to another. One department can serve as a host agency or several departments can combine if they have a regional CAD, for instance.
Smith reported departments are finding that with today’s economy, their budget losses are not being replaced.
They must step back and engage knowledge of their jurisdictions to get the right outcomes for their communities,
looking at the problem people and problem locations. Adding the element of prediction leads to a process maximizing
their time and balancing the generalist and specialist to leverage the best of all that is known.
Chief William A. Farrar, Rialto, Calif. reported that TriiTech’s CrimeView Dashboard assisted their decision-making abilities by correlating data to decision-making, allowing patrol officers to more effectively use their patrol when they aren’t answering calls. “Officers have access to query the data at their level along with over 100 custom queries built into CrimeView Dashboard. These custom queries were set up using criteria that the Crime Analysis Unit (CAU) often follows when responding to various requests by officers, line level supervisors, and commanders.
“Officers can log into CrimeView Dashboard, and within 30 seconds, view repeat locations for robberies or burglaries, broken down by time of day, day of week, and basic information for each incident. Officers apply proactive enforcement measures based on this information when they have free time on their shift.”
CrimeView Dashboard allows decision makers to make some decisions without relying on the CAU. Some basic data requests previously directed to CAU (such as activity by location, call frequency by hour per shift, repeat locations by crime time, violence crime hot spots, etc.) can be answered within Dashboard at the touch of a button. This frees time for analysts to focus on more in-depth research that aides in case clearances, grants, and long-term strategic projects and allows more expeditious decision-making.
Chief Farrar stated, “CrimeView Dashboard includes a feature called ‘Mission’ that allows end users to add information to the system for all users to see (based on user rights). For example, if we are dealing with a crime series in a given neighborhood, an analyst can add a Mission for the event providing detailed crime series info and forecasted results for the next probable hit, including attached crime bulletins, suspect photos, and other data. Officers patrolling the neighborhood related to this mission can add notes on observations, thus providing a sort of log keeping track of actions/recommendations related to the crime series.
“These missions are viewed as a form of predictive analytics but provide more end user functionality/control because the data has been thoroughly analyzed by an analyst/officer to support the predicted future hits. This differs from other popular predictive analytic tools that generate predictions solely on day of week, time of day, location but are unable to account for crime MO and other factors only identified after reading through calls/reports.”
PredPol CEO Larry Samuels stated, “Our solution is designed by law enforcement for law enforcement. The idea for PredPol came from Los Angeles PD when Bill Bratton was Chief. He wanted to know what they could do with data management for the huge amount of data they collected. Bratton went to UCLA and other branches of the University of California and asked them to look at the data and what they could do with it. They determined that with the ‘big data’ they collected, it could be used for predictive analytics.”
Samuels stated that everyone in the community knows about at least two high crime areas and the police probably know 5–7 areas of high crime.
By using a computer with those massive amount of data they could see what patterns emerged. Rather than two or 5–7 high crime areas, the computer can do an analysis and come up with 20 high crime areas.
Using their proprietary patented math algorithm with the known criminal factors, they can deliver data back to police officers in a map-based triangular form. The map will give them a 500’ X 500’ box, of about half a block, with a very accurate prediction. Officers on the street will have a very tight map of what could happen in their patrol area. It is updated in real time and changes shift by shift and by crime types. The triangular box moves by these parameters.
Officers get a briefing at the beginning of the shift but this solution gives them an awareness of what type of crime is most likely to occur during any given day and shift. Officers can spend whatever incidental time is available after answering their mandated calls. The officers make contact with the ‘box’ during their shift whenever possible, driving through, or making community contacts. They might spend 10–15 minutes there during a shift or not at all at times.
This is a long-term systematic effort to reduce crime in the community. Samuels stated, “You are interrupting the opportunity for crime as much crime is opportunistic. Crime goes down because you have interrupted their opportunities to commit crime.”
It is an integral part of what they do, but the process is important. They can show historical analysis to show that their predictions are correct so officers know that the solution will work and that it will make their efforts more effective. They can also see how much time they are spending in the ‘boxes’ and if they spend more time, how effective it will be. They need to know the solution will work and is an effective use of their time.
PredPol has reduced crime around the country and many departments are willing to speak about its effectiveness. For instance, in the Foot Hills Division of the LAPD during the third year of deployment of PredPol, despite a reduction of 25 percent of their officers, there was a 17 percent reduction in the crime rate. “It’s a very effective tool,” Samuels reported, “and it dovetails nicely with current police practices, as well as being easy to use.”
Samuels stated that PredPol is extraordinarily inexpensive and generally the return on your investment is not known for law enforcement activities, but it is available for PredPol. Due to the available off-time for patrolling provided by PredPol, they are able to return 2–5 percent of patrol officers’ time for patrol due to the reduction in crime. PredPol is used in jurisdiction having 10,000 to 4 million people and this can mean multiple thousands to millions of dollars in savings returned just on office time. It also reduces the impact on victims and increases property values. Police departments face more budget constraints than ever before and PredPol maximizes their efforts and puts them ahead of crime.
Samuels stated, “PredPol is the premier solution for predictive policing with the research beginning in 2007. But, predictive policing is actually thousands of years old because we have always known where their problem areas are located. It took world-class technology from a world-class research university to give law enforcement this tool. It is only a tool, but it has a real-time feed from an agency’s RMS that continually affects predictions.”
The results can be printed out, or viewed on a smartphone, tablet, or smart computer. Each officer can get the predictions in the form in which they prefer. There is a version of the solution for any platform.
PredPol does not use personal data to make their predictions. They use the event type, the location, and the time when events have occurred at that location. That is all they need to make their predictions and they have gone up against systems that use personal data and made more accurate predictions without using it. Many police agencies are under fire for profiling and using this solution, it excludes such allegations.
Predictive Policing Is Not Profiling
Departments using predictive policing solutions have not been overwhelmed by complaints about profiling or other concerns. Agencies are not targeting individuals but rather crime patterns and making good decisions to reduce crime.
Chief William A. Farrar, Rialto, Calif. Police, stated that his department has encountered no concerns about profiling or privacy concerns. His department uses TriTech’s CrimeView Dashboard for predictive policing solutions. “All data is safe guarded to maintain confidentiality of incidents. Involved party names (i.e., victim, suspects, witnesses) are omitted to maintain privacy and only names of those arrested are displayed in our CrimeView Dashboard (this complies with government code rules).
“We do not use CrimeView Dashboard to target individuals, therefore, profiling is not an issue. Dashboard is utilized to make informed decisions for proactive enforcement, initiate Problem Oriented Policing projects, identify ongoing crime trends that are impacting the Rialto community, observe year-to-date statistics in order to determine if we are above/below average thresholds for crime, etc.”
Regarding the parameters that might be used for predictive policing, such as offender information, linking traffic stops in drug areas, and police contacts vs. convictions, Chief Farrar stated, “Our Crime Analysis Unit often will take the info in CrimeView Dashboard to conduct a detailed analysis to establish these types of links. This is done proactively by CAU as well as in support of officer requests based on their own searches.
“We utilize calls for service (reactive & proactive), incident report, and arrest data in our CrimeView Dashboard currently. Discussions have come up to add citation data, SRF subject info, and high-profile subjects (such as gang members and violent offenders) along with one of the functions of the NearME component, which would notify officers when they are driving in the vicinity of subjects with active warrants.”
Kathy Marks has been a child abuse investigator for more than 30 years. She teaches classes regarding domestic terrorism and is a previous contributor to LAW and ORDER. She can be reached at firstname.lastname@example.org.