Crime can be predicted – Predictive Policing is here!

“Did you ever wonder how they can discover so much about a criminal’s mind/crime through just data?”

Analytics is the discovery of meaningful insights and communicating it in an efficient way using the data available. The advancement in analytics has led to the development of various techniques which help make people use their resources optimally. Heard of “predictive analytics”? There has been a lot of buzz around this word in the last few years; this emerging area of analytics world can help build models to literally predict future possibilities (i.e. the probability of event happening can be predicted).

Predicting ‘Crime’ with Analytics – How?

Do you know that on an average, out of every 100 people worldwide, 8-12 people are affected by crime related activities like murders, thefts, burglaries, other illegal activities etc.? In India alone, 3.2 crores cases were pending in courts as of 2011 statistics. How those numbers might have grown in the last 3 years is anybody’s guess. Can we stop the occurrence of these crimes or at least try reducing the crime rate? Yes we can; one way is by imposing strict laws which already exists and the other way is to prevent crime. Predictive Policing helps in the latter.

Image Source: Flickr

What does Predictive Policing mean? What does it actually do? Who uses it and for what? How is it related to analytics? Read on for the answers.

Predictive Policing is a method followed by the police/FBI/Detective Agencies to predict the instances of crime. The model is built using statistical algorithm that uses crime data which is updated regularly and a behavioral algorithm that helps in identifying the suspicious human activity. Both these algorithms help in predicting criminal activities.

How does it work?

A time series of data is collected and recorded regularly and this is updated by police officials immediately after the crime is reported. This data will help the police in identifying the hot spots of the criminal activities happening in particular areas. This also provides details on areas in different cities which are at higher risk for repeat offenses. The behavior algorithm is used for identifying unusual activities (like sudden movements, trespassing in restricted areas, leaving luggage in crowded areas etc.) of individuals in crime prone areas.

To understand the concept better, let’s look at a real-life example – ‘PredPol’ is a software (used by Los Angeles Police Department (LAPD)) developed by mathematician George Mohler and anthropologist Jeff Brantingham. The software is similar to the one used by online retailers Amazon and Wal-Mart to collect data on customers’ buying habits to predict buying behavior.

How analytics is used

Predpol, based on its algorithm, generates maps that display 500 x 500 square-foot hot spotswhere crimes are likely to occur again. The algorithm uses criminal records, date and time of crime occurrences, location and many attributes to predict the crime prone zones. The algorithm identifies the patterns of criminal activity based on past data; maps are generated based on the shifts as the frequency and probability of a crime occurring, changes with time. The hot spots help in deploying more staff in that area, so that there is better preparation for prevention of criminal activities.

Omniscient cameras are used for surveillance purposes; these are pre-programmed with a list of normal human behavior. Once they detect any unusual activity of individuals, cops are alerted via a call/text message. “Predictive Policing” methods such as these have helped reduce crime (mostly in activities like thefts, home burglary and suspicious arrests) by 25%, as reported by LAPD police and patrolling.

Source: http://www.theomegagroup.com

Are all these predictions true? May not be a 100% of the time, but it still aids prevention and keeps system alert. While predictive policing works on the basis of past data and predicts future possibilities, it does not replace the police intelligence. It however facilitates the process and helps them track crime prone areas better. The goal of predictive policing is not to increase the number of arrests or cases lodged; it is to deter criminals from committing crimes.

What can we do using crime data?

We can use the data and build models to find key drivers of criminal activities. e.g. person’s criminal record, family background, age group and many other aspects. However you can derive a better probability of criminal activity after making models based on different parameters. For example we can do regression, classification and clustering models to find the hot spots and probability of criminal activity to be deduced. Risk terrain analysis can be used for identifying geographical locations with a high risk of crime.

Predictive policing leverages advanced analytics to enable information based approach towards law enforcement tactics, strategies and policies ensuring public safety. This methodology can be a boon during times of recession; In ares with man power shortage, this system can prove to be a cost benefit by identifying hot spots and predicting crime prone areas.

Method Adoption:

• Santa Cruz Police Department adopted predictive policing and found crime rate falling down.
• FBI extensively uses predictive policing, in fact they were one of the first to experiment in small populous areas to find the effectiveness of models. They are still using these approaches to solve cases (In 2011 they solved a murder case committed in 1978, by gathering finger prints and matching it with probable suspects, leading to the arrest of the murderer).
• Few American States followed this system and found it to be useful in reducing the crime rate mostly in theft, burglary and illegal activities (like smuggling).
• A similar software developed by IBM for the South Carolina police has also led to a decrease in crime.
• In India, the Jharkhand police are trying to implement this system with the help of IIM Ranchi, by building sophisticated algorithms and behavioral science which compiles historical crime related data.

With the advancement in technology and adaptability, Predictive Policing might become an important tool in law enforcement and crime prevention.

Challenges

• Can a suspect be arrested and produced in court of law based on the prediction by a computer system? How can that arrest be validated? How effectively can police use this system to prevent crime? It also demands a great level of honesty and integrity from the police department.
• If criminals identify this system, they may move towards the less crime prone zones and this approach might put public safety in jeopardy.
• We don’t know how much it would cost in future and officials willingness to spend on acquiring best methods for crime prevention using surveillance, behavioral, statistical modeling due to advancement of technology.