HYDERABAD: Blending Sherlock Holmes’s logic and intuition with modern technology, the National Crime Records Bureau (NCRB) is attempting to use crime data analytics software to enable predictive policing so that crime can be nipped before it happens. By March 2018, predictive policing technologies will be available to five states -- Kerala, Odisha, Maharashtra, Haryana and Tripura -- and by the end of 2018 to all states and union territories.
Sources disclosed to New Indian Express that software using crime data analytics is being developed by the Hyderabad-based Advanced Data Research Institute (ADRIN) so that laws enforcement agencies can map crime patterns, conduct hotspot analysis and predictive analysis.
"It is basically use of big data analytics to predict geographical areas where crime is most likely to take place, after taking into account various factors followed by analysis. Police can then step up patrolling, deployment and surveillance so that offences do not take place. The NCRB recently signed an MoU with ADRIN, which is working on the software. It will be ready by March 2018," a source explained.
"Not only offences, it will also go a long way in predicting terrorist and Maoist attacks as well," said NCRB director Ish Kumar.
One of the key features of the ambitious project is Hot Spot analysis wherein police personnel will analyze and predict geographical areas of increased crime on the basis of crime data. Hot sport analysis will process all patterns relating to time of occurrence, exact locations, even shops, hotels, bars or other establishments.
'Detectives will study the social and infrastructural environment of a location and its local demography. Analysis will also be made of such things as the local weather conditions, the days on which offences are committed, say whether it was around pay-day, etc. By analyzing all these factors, we can predict whether offences will be repeated around the same time, same weather conditions and same days.
“It is a prediction. One cannot say it will be 100 per cent accurate. But in countries like USA, predictive policing has given great results,'' explained an official.
Based on big data, sleuths will analyze other factors like 1) which are the communities living in the particular area, (2) their population, (3) their lifestyle, (4) general behavior, (5) social issues -- all of which can be useful in assessing whether a particular area is crime-prone or not.
"Crime pattern identification, crime prediction and alerts, redeployment of police resources, monitoring social media and news analysis relating to crime is part of the project," the expert said.
But the key to the success of the project is the availability of historical crime data for which too software is being developed. ''Some countries have crime data going back 70 or 80 years. The NCRB is also planning to have something similar,'' the source said.
Predictive policing is like a crime forecast involving algorithms, just as it is done in predicting earthquakes. Data mining techniques will be used to recognize patterns in the data and use the same to make predictions. Different algorithms are used to mine data in different ways and the most commonly used are clustering algorithms and classification algorithms. The software will use a certain algorithm which will predict when and where a crime is likely to take place.
“But the key to it is the massive data on previous crimes and how best our people are able to analyze and correlate them with the present crimes," the official said.
The NCRB’s Ish Kumar said countries like USA have used the same technology and have been able to reduce the number of offences. ''We have obtained data from states like Kerala, Maharashtra, Odisha and Tripura to develop models for predictive policing and once the tools and models are finalized, these states will have the facility in the first phase. Subsequently, it will be released to all states in a phased manner,'' he said.
How they did it
The Los Angeles Police Department reduced burglaries by 33 per cent, violent crimes by 21 per cent and property crimes by 12 per cent by adapting earthquake prediction software to crime prevention. Thirteen million crimes committed over 80 years were fed into the mathematical models to predict areas where crime was most likely to occur. New crimes are constantly added to the database.
The Chicago police have gone beyond hot spots to using bid Data to target people most likely to commit a crime in future.
The New York Police Department started using crime data about 20 years ago. To keep track of crime trends, NYPD used the software CompStat to monitor crime trends on a daily basis.