Sunday, November 13, 2011

Location Intelligence for Bottom-Up Criminal Gang Analysis

Last week I visited a crime analysis center in the mid-Atlantic region.  I was there to provide consulting on enterprise GIS design and implementation.  The biggest challenge at this center was processing of incoming data -- automated entity extraction, normalization, and geocoding.

But I was also interested in their analysis process and how GIS and location intelligence can contribute.  Gang crime analysts at this center spend a lot of their time trying to categorize crimes and arrests by gang.  That is, labeling the suspect or event as belonging to MS-13 or Bloods or Crips, etc.  At the same time, they admitted that their partners and stakeholders don't always agree on the attributes of each gang, or even what is a gang-related incident or person.

I wondered, Why is it important to assign a gang label?  Is it so that similar behaviors, attributes, and territories can be assumed for all persons and events assigned to the same gang?  So that law enforcement can exploit contacts and leads already established within a well-known, named gang?  The best answer from the gang analysis team was this:  Categorizing helps law enforcement agencies allocate resources.  (Translation?  If you have an MS-13 or Bloods or Crips problem in your jurisdiction, you get more funding and other resources?)

I thought that a different approach might lead to better interdiction:  Instead of assigning gang labels from the top down, work from the bottom up.  Like this:

Use traditional social network analysis (ie link analysis) to define and diagram links between individuals and events, using whatever linkage data isavailable -- phone calls and address books, crime and arrest locations, last-known address, relatives and associates.  Most of these data inputs have a location component.

Courtesy of cid.army.mil .

Use geographic proximity, along with more complex location and societal algorithms, to calculate relative weights for each link between two individuals.

As the network or "gang" grows outward from a single person or crime/event of interest, give it an arbitrary name.  Forget about MS-13 or Bloods or Crips for now.  If an association with a big-name gang emerges, that is a useful data point.  But it's not the focus of analysis.  The focus is to understand the close-in, immediate gang or family or mafia that more certainly characterizes the suspect or event of interest, and which can be exploited.  Rather than boxing the group into big gang categories.

I'm sure that most of the crime and link analysis software packages can do weighted link analysis.  Not sure, though, how they integrate location into their formula.


No comments:

Post a Comment