Predictive Policing is the use of analytical techniques and tools to identify and ultimately prevent potential crimes before they ever happen. It uses large amounts of data about past crimes to find patterns and trends, then use that information to improve patrols and make the most of police resources.
Predictive Policing is Powered by ML and AI, machine learning and artificial intelligence are likely to become invaluable tools for law enforcement.
“It is all going to come down to artificial intelligence, because no analyst can sift through the sheer volume of what investigators have access to and need,” said Cathy Lanier, former Chief of the Metropolitan Washington, DC Police Department in The Changing Nature of Crime. “Artificial Intelligence will become 100 times more effective than a person in sifting through and narrowing down data to what is important.”
One of the most promising applications of ML and AI is in predictive policing. Software offering analytic capabilities like Visallo, automatically link people, locations, objects and events, allowing law enforcement to uncover hidden relationships across very large data sets.
But does it work? According to one study on the impact of predictive policing on robberies in Milan, Italy “predicting policing improves police patrolling by a large degree… benefits appear to outweigh costs by a factor of 5.
Interested in how data visualization tools can help find hidden connections in your organization’s sea of data? Sign up for a demo of Visallo to talk to our team.
Thinking about diving in to a data-centric model for predictive policing? Here are a few things to think about as you get started:
Combine new technology with time-tested investigative techniques - Very few crimes can be solved with technology alone. Many software companies position themselves as an end-to-end solution. At Visallo, we know that the human remains the most important factor in the predictive policing process. Even with the most sophisticated software, law enforcement is still required to choose which data to analyze, ensure the data is high quality, design the analysis, interpret results, and decide what action to take. Visallo was developed to augment, not replace, the human.
Predictions need to be assessed by experienced law enforcement officers - Technology isn’t a crystal ball. Investigative link analysis software like Visallo can help law enforcement assess the risk of crime, but it can’t predict whether a crime will occur. Visallo makes it easier to identify patterns, but it still relies on the experience and expertise of law enforcement to know where to look. Visallo also makes it easier to share insights with colleagues and collaborate across teams.
Use high quality data - There are three main causes of poor quality data: censoring, systematic bias, and relevance. Censoring is omitting crime data from a particular location or time. If data is censored it will appear no crimes occurred. Systematic bias can result from how data is collected. For example, if the data shows a spike in home break-ins from 7- 9 am, that may be because victims wait until morning to report the crime. Relevance describes the usefulness of the data. For some investigations, it is helpful to have data going back many years or even decades. For investigations involving a recent crime spree, this amount of history may be less relevant. Law enforcement is needed to make sure data is high quality. Visallo syncs directly with law enforcement data and updates automatically.
Sometimes simple models are the best models - Keep in mind that model complexity is not necessarily correlated with predictive accuracy. Many times, simple heuristics are just as predictive as complicated algorithms. Visallo’s intuitive visualizations help law enforcement uncover hidden relationships and connections, without relying on overly-complicated black box algorithms.