In the last couple of years, one of the most interesting application areas where machine learning has found traction is crime prediction. City police departments, across the globe, have been enthusiastically embracing the technological intervention that can improve crime analysis or its prevention rates. By focusing on crime prediction, police forces can not only focus on increasing optimization and efficiency in police force deployment but also prevent crimes. While we are far from the science fiction of Minority Report becoming a truth, various reports have suggested that implementation of these software has helped police departments in either preventing or accurately predicting crimes by approximately 10-15%. This is a significant number, given that we are talking about saving human lives at risk here. Leading the effort in this field have been startups that have specifically focused on building cloud based predictive policing solutions for the police departments. In today’s post, we will look into some of these machine learning startups that have realized the dream of predictive crime prevention into reality.
Built by Philadelphia based Azavea, HunchLab is a cloud based predictive policing software that claims to predict the next crime scene, incorporating factors ranging from events, terrain, climate, routine activity, etc. Its machine learning algorithm can also probe for near repeat patterns, aoristic crime analysis and risk terrain modelling. Miami police uses the software.
Santa Cruz, CA headquartered PredPol is a predictive policing software for better police patrol management. PredPol software predicts crime prediction probabilities for any region under its purview. The tool isolates each region into small areas, depicted in 500 feet by 500 feet boxes on maps, and determines its crime probabilities. This data is refreshed in sync with the patrol shift timings. Armed with it, patrolling can be optimized and strengthened in areas that are more vulnerable to criminal activity. Its customers include the likes of LA Police, Atlanta Police, Santa Cruz Police and Kent Police (UK) among others. The software also claims that crime rates, after its deployment by the local police forces, have reduced by 10-30%.
Information Builders – Law Enforcement Analytics (LEA)
New York based Information Builders has built its LEA software as a comprehensive dashboard solution for police departments and units. Using it, the police department can generate real-time analytics on all its historical data, including keyword based searches on structured and unstructured data. The tool also has a crime predictive tool and social media data integration into its search. Charlotte-Mecklenburg Police Department (CMPD) is one of its customer. Bair Analytics and Datameer are among other players that offer similar solutions. Bair Analytics solution also has a freely accessible crime data component for the residents.
Motorola Command Central Predictive
Accessible from any mobile device, this cloud based solution claims to accurately predict 30% of next day crime locations and types. Using predictive analytics on historical data, the system claims worthy of being used in optimizing police force movements and prevention of crime. St Louis Police Department now uses this system. Motorola suite also has complementary solutions such as Real-Time Intelligence Client that aggregates and correlates data from multiple sources such as video, sensors, alarms, etc. and improve the situational awareness of the police force.
Similar to Motorola, Hitachi also revealed a similar product called Hitachi Visualization Predictive Crime Analytics (PCA) suite that blends the power of predictive analytics and visual sensors layered across the city for real-time crime predictions.
The Milan Police Department uses the Milan-based KeyCrime software. KeyCrime employs analyzing criminal behavior analytics data to predict crimes. The software captures each bit of tiny information about a crime scene, the location, the suspect physical attributes, the weapon and crime vehicle details, police reports, victim narratives, visual evidence and analyses all of it to build predictive criminal strategies. The output is in the form of a report that contains suspect future target profiles, physical attributes, his modus operandi, possible transportation he might use and the date and time range predictions.
Series Finder – MIT
In 2013, Cambridge Police Department (CPD) and MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) jointly worked to develop a modus operandi constructive model for crime offenders. By analyzing historical data on break-ins (door type, week day, property and neighborhood characteristics), the system was successfully able to identify several previous crimes, undetected by the local police, with identical crime pattern patterns.
In addition to these, numerous other police departments are directly engaging with predictive analytics service providers to build their customized crime analytics and prediction tool. Established players such as Microsoft (Domain Awareness System) and IBM (Predictive Analytics Lead Modeling Software) are also operating in the market, helping police departments with predicting policing solutions. With time, as more and more crime data is aggregated and analyzed, the prediction rates of crime types will increase further. And while these systems might not be able to predict all the crimes, they may go a long way in helping the police force significantly control incidences of actual crime.