Solving crime through big data analysis

Pooling crime statistics, social, demographic and geographic data enables analysts to speed up crime solving and even help predict where crime will occur, argues Ryan Prox of Vancouver Police Department. James Atkinson reports

Solving crime through big data analysis

Intelligence-led policing is much talked about as a future goal, but in some places it is already a reality. Vancouver Police Department in Canada is not only analysing data to help solve crimes, it is also using it to predict where they are likely to take place.

As Ryan Prox, Head of Analytic Services (pictured), Vancouver Police Department, explained to attendees at the British APCO 2015 event in Manchester at the end of March, the impetus behind this has come from both the changing nature of policing and the use of new technology.

‘What used to take years, public safety agencies can now do in months, and police offers are demanding the same technology for work that they can already access privately,’ he says.

‘Vancouver Police has an extremely advanced and tech-savvy management, but more importantly its new officers are weaned on technology from birth. That changes the way they look at things.

‘They are very demanding in what they want and how they get it. They expect to see more advanced technology in police forces than they see in everyday life, but in many cases it is the exact opposite.’

Holistic approach
Prox argued that it is important to take a holistic approach to public safety issues and to look at and try to address their root causes.

This involves providing the police with social demographic information on communities at risk, including data on community infrastructure, school resources, mental health programmes and housing assistance resources.

‘We need to drill down into the core causes that are driving crime and deliver it to officers at their fingertips. Teachers can spot potential criminals from the age of six. They are better at identifying them than any software we can develop, and it means we may be able to put in intervention resources at an early stage, which may stop them going sideways,’ says Prox.

The above information then needs to be combined and mapped against victimology patterns, environmental factors and other statistics, such as detailed representations of police activity and responses to public safety incidents in real time. This enables the police to see what is going on, assess the situation, and then adapt or change their response if required.

But to be successful Prox emphasises that it is necessary to engage public, private and academic partnerships. He acknowledged that these three sectors may have competing agendas. ‘But you need this tri-partnership to drive change and technological innovation.
That way you get new things and you are not responding to past events; instead you get what you need now.’

Police forces need to cultivate an intel-led environment, he argues. ‘All police departments say they are intel-led,’ says Prox, ‘but if you have 2,000 officers and just two analysts, are you really evidence based intel-led then? I’d say no.’

Organisational objectives
Prox argues that police have to come up with some organisational objectives and then align their people, processes and technology to support intel-led policing. Key goals include: supporting advanced analytic processes; empowering police members to be proactive; and enabling an ability to target crime faster and smarter, which saves time and therefore money.

Failure to align people, process and technology means that forces are either chaotic or reactive, instead of being proactive, according to Prox. Technology needs to be used to simplify processes, making officers more efficient and effective. An integrated resource management system (RMS) platform combining patrol details, intelligence and major case management is also required

Silos of information must also be eliminated so that different repositories of data can be cross-referenced and ‘talk to each other’. Prox cited the example of mass murderer Robert Pickton, who was finally convicted in 2007 after killing a possible 49 women.

‘We had the information that would have led us to him, but no one connected the dots. If we’d been able to confederate the data we’d have got him earlier,’ says Prox.
Records therefore need to be consolidated to enable an easily accessible intelligence data mining environment. What’s needed, according to Prox, is a ‘Walmart of data repositories’.

Data warehouse
The Vancouver Police Department data warehouse holds information such as: court disposition data; municipal human source repository; arrest and booking information; social network analysis; geospatial information, including social demographics and critical infrastructure; police briefings; and an offender MO database – on sex crimes, for example, built up from templates filled in by police officers. Real-time CAD data is also saved every 30 seconds and fed into the database.

All of which provides an integrated analytics database, which the police analysts can use and send out to officers in the field. Prox recommends that commercial off-the-shelf products should be used to manage and access information.

‘COTS vendors support loads of other users and they will keep you up to date because you are using common products – that way you will not be stuck on a legacy system,’ he points out.

Vancouver Police uses: Microsoft’s SQL server for analysis, reporting and integration services; IBM’s i2 Analyst Workstation/Notebook software suite and ESRI connector; ESRI’s arcGIS server, which integrates into i2 data mining and query tools and with Analyst
Notebook and desktop.

VPD also deploys Versaterm’s Versadex records management system and simplified data integration; police computer aided dispatch (CAD) and mobile data terminals; Citrix for remote access; and Latitude Geographics Group for mapping services.

Integrated analytic findings can be displayed visually on the analyst notebook chart. For example, by inputting crime incident data, offender MO categorisations and social data, an analyst can see how a particular subject is linked to other gang members and persons by presenting a vector chart of their associates and so on – and all in 30 seconds.

Prox reports that VPD has a sex crime team of 40 people, but once an analyst was attached to the unit and able to look at the MO characteristics of offenders listed, the analyst cleared more files than the entire sex crime team combined on a regular basis.

‘The way it works now is that the on-scene team gets all the details. A sergeant calls the analyst who works on the MO detail and pulls up suspects from the database. The police officers then go and look for those people. The clearance rates have tripled to quadrupled using this new model,’ says Prox.

For example, Ibata Noric Hexamer, a rapist and child predator, was investigated by VPD for a year without results. The force was unable to match DNA from the crime scenes to any of the 561 known sex offenders listed on the police database.

The VPD analysts then had a look at the material and came up with a suspect who had no criminal record, but by analysing other information, such as housing sale movements, were able to come up with a suspect in just five or six weeks.

A surveillance team was put on Hexamer and DNA was collected from a discarded coffee cup and matched to DNA collected from the crime scenes – it proved to be a 1-in-6 quadrillion match.

The success of this case helped sell the approach to VPD. ‘We replicate this process on a day-to-day basis now,’ says Prox, ‘We look at the patterns and the MOs.’

VPD now has a live law enforcement situational dashboard called GeoDASH on a cellular connected mobile data terminal (MDT) in every police vehicle and office desktop. The police are using Panasonic Toughbooks, although the interface had to be redesigned with bigger buttons, as the original ones were too small for use in a car driven at speed.

GeoDASH empowers VPD members to be proactive as it is a patrol-driven project, which supports intel-led policing by providing situational awareness and instant access to relevant resources. This enables the police to target crime faster.

Predictive policing
But it does more than that. It can also be used to forecast the movement of crime. Algorithms are used to look at crime records, such as historical break-ins, to predict where they are likely to happen.

Other information such as LiDAR data (showing if a neighbourhood has a high occupancy or not – if yes, it is less likely to be targeted for theft), residential property sales, commercial uses, traffic congestion, weather, time of day and so on are all fed into the model.

Predicted crime maps can then be created and displayed on the mobile terminals. Prox says the display shows what incidents are likely to happen over the next eight-to-12 hours with 78% accuracy.

Areas are divided into 100m by 100m grid cells and low, medium and high crime predications are relayed every hour. ‘This allows officers to site themselves near predicted crime scenes and make arrests when the crime is in progress,’ says Prox.

VPD is continually evaluating the model and examining the best response. For generic crime the response is generally to increase patrol resources in an area. If it is crime specific they conduct forecast crime specific interventions.

If they are attempting to handle a problem-specific issue then a wider, more holistic approach might be called for to address a particular location or the factors driving crime-related problems.

So, if it is a socially and economically challenged area, the police can work with social services to set up centres where local people can go. Crimes such as vandalism often then drop off.

‘There is a causal relationship between the information being collected and what gets sent back out into the field to help officers and so that encourages officers not to do a sloppy job,’ observes Prox.

Looking ahead, Prox says 4G mobile networks are enabling assets such as public CCTV to be used more efficiently. But feeds from cameras used by security and building management companies can be pulled in too.

If the police know where the private CCTV cameras are located, that means less video canvassing to try to find footage after a crime, as they know where the cameras are and what they are covering.


Leave a Comment