It is afternoon. Mandeep Singh Randhawa has just arrived at his office, flanked by two junior officers. Around a dozen men and women wait patiently in a room next to his cabin. A few certificates of merit adorn the walls.
“Please give me a few minutes,” Randhwa, deputy commissioner of police (Central Delhi), tells the visitors and moves straight to his seat and reaches out for the desktop. As he moves the mouse, the monitor, which is on the sleep mode, comes alive with a colourful geospatial map of the area under his control. He zooms in on a micro-scale section of the area, which has crime hots pots marked in red.
Randhawa applies a few filters on the new crime-mapping software, and examines the calls made to the police control room about street crimes such as snatching, robbery and vehicle thefts in his territorial district. Pausing for a few moments, he reaches for his wireless handset and connects all ACPs and SHOs of his district one by one.
“Deploy two bike patrolling teams on either side of Bahadur Shah Zafar Marg in the press area. There were two snatching and robbery calls between 7 pm and 1 pm last night from the stretch. I don’t want a rerun of the crimes tonight,” he sternly tells the SHO of the IP Estate police station over the wireless set before scrolling through the application again.
The DCP has identified the crime hot spot and possibility of any repeat of the crime in the area through a new software called CMAPS (Crime Mapping Analytics and Predictive System). The web-based software accesses real-time data from Delhi Police’s Dial 100 helpline and, using ISRO’s satellite imageries, spatially locates the calls and visualises them as cluster maps to identify crime hot spots.
The software’s statistical models and algorithms help the police understand where the next crime is likely to occur. This concept called ‘predictive policing’ marks a paradigm shift in policing in the national Capital spread across 1,483 square kilometers and divided into 13 police districts, each headed by a DCP. Police get over 27,000 calls on Dial 100 every day on an average.
“We are not predicting the future, what we are predicting is where and when a crime is likely to happen based on scientific and objective analysis of data. And this will determine where our next observations are going to be,” says Sanjay Beniwal, special commissioner of police (operations) sitting in his ninth-floor office at the police headquarters, which gives a panoramic view of the city. “In a mega city, we believe a lot of crimes are predictable and can be curbed,” he says.
CMAPS, which updates the data every three minutes, replaces Delhi Police’s mechanical crime mapping —which involves manually gathering data at an interval of 15 days.
The new software, Beniwal explains, allows the information generated though Dial100 to be plotted on the geospatial map of Delhi, enabling the police to pinpoint exact locations and spatial distribution of the crime. “We can study the pattern of calls coming from a particular location. Suppose if there are too many robbery calls coming from a location which has a bar nearby, we can match those calls with past crimes in the area and understand if that bar is in any way responsible for these robberies, and whether another crime is likely to occur there,” he says.
Many cities such as Los Angeles and Kent in the US, and Berlin in Germany are using predictive policing techniques to prevent crime. In fact, the idea of crime forecasting took shape in 1931 when two Chicago criminologists, Clifford R Shaw and Henry D McKay, conducted a research, exploring the persistence of juvenile crime in some neighbourhoods of the city. They came up with the social disorganisation theory, which states that location matters when it comes to predicting illegal activity.
That is when scientists began experimenting with data technologies, including statistical and geospatial analysis, to determine prevalence of crime. The researchers have been using everything from basic regression analysis to cutting-edge mathematical models to predict when and where the next crime is likely to occur. But until recently, the limits of computing power and storage prevented them from using large data sets.
What softwares such as CMAPS have done is allow larger data collection, and process it in a much more sophisticated mathematical way — a significant improvement over just manual hotspot mapping, which Delhi Police have been doing so far.
But statistical model-based predictive policing has raised concerns in the West that it might create prejudices against particular communities or areas and how much power should be accorded to computer algorithms.
“Any statistical model can be flawed, but we hope it will be too dynamic to hold any area guilty of a crime for long,” says Beniwal.
“The web-based software will soon be available on mobile and even a beat constable will have maps showing locations where crime is likely to occur,” says an officer. “It is all about using science to curb crime,” says Beniwal.
WHAT IS PREDICTIVE POLICING
- Predictive policing is the application of data analytical techniques to identify likely targets for police intervention and prevent crimes by making statistical predictions. A crime forecasting software involves statistical models and algorithms
How it works
1 Predective policing starts with collecting large amounts of data on past crime and co-relating them with present crimes. CMAPS, Delhi police’s crime mapping and predictive software is integrated with Delhi Police’s new software and allows the information getting generated though Dial100 to be plotted on geo- spatial map of Delhi
2 Analyze Data: Software looks for patterns and correlations in past crime data
3 Predictive Maps: Algorithm predicts where and when a crime is likely to happen in future
4 Increased surveillance: Based on the possibility of the crime, police add or redeploys resources during certain period to prevent crime