To Predict and Serve: Predictive Intelligence Analysis, Part I

Posted on 5 July 2011 by


By Sgt Christopher Fulcher

In law enforcement, as in major military missions, the greatest weapon against constantly evolving threats is the intelligence that we gather and then use to drive strategic decisions. Even with some of the best intelligence collection in the world, the process is largely a reactive one.

The goal of our intelligence activities is to expose security threats in time to take action. Individual acts, though, are usually hidden within enormous amounts of data which must be analyzed and countered before the next event occurs.

With so much information to collect and analyze, useful data is sometimes overlooked or not considered at all.

This has made many local agencies data rich but information poor.

What if this reactive process could become proactive? What if intelligence could be analyzed in such a way that the result provided measurable and actionable information about events before they occurred? What if you could shift the intelligence paradigm from “sense, guess, and respond” to “predict, plan, and act”?

Technologies and methods on display in the 2002 movie Minority Report look far less outlandish than they did nine years ago. Predictive analytics is now a valuable and useful intelligence tool for law enforcement that, when used appropriately with other intelligence methods, will produce a positive impact on local crime and domestic terrorism.

Predictive analytic systems have proven successful in anticipating threats, identifying suspicious actors, and effectively allocating resources. They are valuable for identifying concealed patterns within large amounts of structured and unstructured data.

Strawberry Pop-Tarts and Beer
You are more familiar with the concept of predictive analytics than you may know. Predictive analytics was born in the business world and has been used successfully for many years. Retail stores, sales, marketing, and even cell phone providers’ “churn” models are just some examples of predictive analytics or “data mining” at work.

Models in the private sector that are used to analyze shopping patterns and examine purchasing decisions can also be used by the public sector to identify a motive or predict the next incident in a series of crimes.

In 2004 when hurricanes were damaging the southeast, Wal-Mart decided that there was a competitive advantage to the destructive weather pattern. Wal-Mart’s chief information officer began to data mine sales data and found that strawberry Pop-Tarts increase in sales, approximately seven times their normal sales rate, ahead of a hurricane. They also found that the pre-hurricane top selling item was beer.

Ever purchased music from iTunes? Rented a movie from Netflix? Made a purchase with a credit card? All of these everyday tasks produce detailed complicated mathematical algorithms behind the scenes to determine the popularity of your choice.

These algorithms are so advanced and powerful today, they can predict what you might want next and offer suggestions you may have not even knew you wanted. Look no further than the iTunes Genius application. James Guszcza from Deloitte states flatly that, “Data will be explicitly recognized as an asset”.

Data Mining
Data mining is a process used to discover hidden patterns and relationships in large amounts of either raw or previously analyzed data. The process of analytic technologies can help quickly and efficiently identify actionable pieces of information from within larger data sets. This becomes a fluid process as the analysis can be accomplished even as more recent data is collected and added.

As the process continues, it provides for accurate and reliable predictions of future events by drawing from the characterization of patterns and trends in historical data.

One of the inherent benefits with a predictive analytic solution it is not necessary to know exactly what you are looking for before you start. Analyzing such large amounts of data in such a proficient manner will often reveal something new that may not have been seen previously.

This is referred to as a “data driven” or ” bottom up” approach because the process begins with data and ends with theories which are built based on discovered patterns or trends. This data driven method is where the true success and “forecasting” of predictive analytics comes from.

A data driven model becomes more proactive than the traditional “top down” method of analysis, which draws hypotheses first and looks for data to support them after the fact.

Domain knowledge is a critical skill when using any predictive analysis solution. Domain knowledge can also be considered “corporate” knowledge and is the art of knowing your field and being in a position to constantly evaluate the validity of your results.

There is no one better than a local representative to know an area and understand the value of the results that data mining can provide. This is one of the things that make data mining so valuable to local organizations.

The Crime Pyramid
Predictive analytics as a crime or terror fighting solution does not require advanced degrees in statistics or expensive software packages. The availability of tools specially designed for law enforcement use is growing every day. In addition, data mining techniques can be applied with existing data sets and little software intervention.

If we look at this data as an inverted pyramid, similar to the old food pyramid, we can easily see patterns take shape. For example, crime rate is often a metric that local agencies use to determine the effective of their department, socio-economic status, and the perception of safety, but “crime” is a large category. Violent crime is a subset of all crime, but is still too large to usefully develop any models or trends.

Murder is classified as a violent crime but can still be attributed to many different motives. A robbery related murder is likely very different than one that is domestic related.

At this level predictive analytics can become useful. How many domestic related murders have occurred? Where have they occurred? Now that a useful association has been developed, accurate and reliable models can be built to visually compare these relationships.

Victims previously involved in aggressive or violent patterns of crimes may be more likely to be shot if they were also known to carry a weapon, which might be related to aggressive patterns of behavior.

Now another relationship of victims to gun permits has been revealed: A crime map showing recent domestic disputes and subjects with gun permits may show a predictive relationship of future crimes.

Next: Blue CRUSH and other analysis programs

Christopher Fulcher is a Sergeant with the Vineland Police Department in Vineland, NJ. He has worked for the police department for 15 years, the last six as a supervisor and the last four assigned to the Services Division functioning as a Chief Technology Officer. Fulcher has been involved with his department’s technology and network management since 2001. Through several educational seminars and classes he has also recently begun tactical, strategic, and geo-spatial intelligence analysis for his department. In April 2011 he graduated from the US Department of Homeland Security Intermediate Fusion Center Analyst Training (IFCAT) program.