Skip to main content
Icons of Progress
 

Predictive Crime Fighting

 

Information analysis is the brains behind public safety. The goal of intelligence activities is to uncover security threats in time to take action against them. But the patterns that point to these threats are often hidden in massive amounts of data. To meet this challenge, one form of information analysis, predictive analytics, is particularly useful.

Predictive analytics solutions apply sophisticated statistical, data exploration and machine-learning techniques to historical information to help agencies uncover hidden patterns and trends—even in large, complex datasets. By using predictive analytics, you can better anticipate what types of intervention will be needed, and where. So you can plan, rather than react. And make the best use of available resources.

How it works

Predictive analytics can seem a bit abstract. Here’s a three-step description of how predictive analytics connects hard data into real action.


1. Connect all the data
First and foremost, all available relevant information must be collected. And we do mean all. For example, lots of information is considered “unstructured” because it does not exist within an established data model. But that information is still important, and sometimes it’s the most important. Unstructured data includes emails, text messages, audio and video files, health records, journals and open-ended survey responses. One of the strengths of predictive analytics is its ability to “structure” this text so that it fits into an analytics program and becomes a piece of the puzzle.


2. Turn that data into insight
Next, a “normal” behavior must be defined and described for all this data, such as normal movements and normal instances. By knowing what “normal” looks like, unusual activity and deviations from the norm are readily identified as they stand out against a backdrop of normalcy. There are three general approaches:
    a. Prediction explores possible relationships and patterns in historical data, determining which combination of factors, such as behaviors and characteristics, are most likely to result in a particular outcome.
    b. Association identifies events that occur together and, given a series of events, determines what action is likely to occur next.
    c. Clustering finds naturally occurring groups in data that exhibit similar characteristics.


3. Present insight in actionable form
Insight gained from predictive analytics is valuable only if it is accessible and easily understood by the people who can act on it. Solutions such as graphing, mapping and statistical modeling can deliver results clearly and cost effectively, often in real time.

In contrast to rules-based analysis and detection methods, predictive analytics can identify relatively unusual behaviors, even those with subtle differences that other methods can miss. Predictive analytics techniques explore and learn from all dimensions of data, thus allowing analysts to combine human knowledge, first-hand experience and intuition to guide the application of analytical techniques. Because predictive analytics are able to combine a wide variety of data dimensions, types and sources on an ongoing basis, it is possible to quickly and reliably detect inadvertent signatures from criminals.

The IBM z/TPF operating system

The IBM z/Transaction Processing Facility (z/TPF) was originally designed to handle the transaction-heavy workloads of airline reservation systems. Today, z/TPF has evolved to provide high availability for the most demanding and high-volume, real-time transaction processing of mission-critical applications.