An Analysis of Teen-Age Driver Crashes 2005-2008 (pdf)

  January 10, 2010      Analytics, Motor Vehicles, Traffic Safety
Brown, D., A. Watkins, CAPS Research Report, Jan. 10, 2010.

This study was conducted at the request of an advocate group that wanted information to assist them in developing public information and education countermeasuers for teen-age drivers. While most past CAPS studies of youth-involved drivers were limited to 16-20 year olds, the advocate group was also interested in 15 year olds, and they were not interested in 20 year olds since their projects were oriented around teen drivers. Several studies were conducted, including CARE IMPACT studies that compared 15, 16-19 and 15-19 year old causal drivers with causal drivers in the older age group.

CARE Driver Distractions Study (pdf)

  November 21, 2009      Motor Vehicles, Traffic Safety
Brown, D., CAPS Research Report, Nov. 21, 2009.

This was a recent report requested to provide information for the Distracted Driver Summit held at UAB on December 3, 2009. A variety of driver distraction causes are discussed and compared.

A Relation Context Oriented Approach to Identify Strong Ties in Social Networks

  October 1, 2009      Analytics, Law Enforcement
Li Ding, Dana Steil, Brandon Dixon, Allen Parrish, David Brown; Annals of Information Systems, Oct. 2009

Social network graphs have been found to be an extremely effective tool in the identification of potential perpetrators of criminal activity. These graphs can grow extremely large, as illustrated by an example within this paper that contained over 4.9 million nodes and over 211 million edges. Obviously some reduction of these graphs is essential to their being useful. Further, considerable "noise" (false positive relationships) are generated when the graphs are totally comprehensive. This research transformed the original social network into a relational context-oriented edge-dual graph. This was done by evaluating the quality of the connectivity for each edge to obtain a metric to this effect for each edge. By retaining only the strongest edges the overall graph becomes more reliable and more useful in practice.

PerpSearch: An Integrated Crime Detection System

  June 8, 2009      Health and Human Services, Law Enforcement
Li Ding, Dana Steil, Matthew Hudnall, Brandon Dixon, Randy Smith, David Brown, Allen Parrish, IEEE International Conference on Intelligence and Security Informatics, Dallas, TX, Jun. 8 - Jun. 11, 2009.

This paper presents the first attempt to integrate four distinct approaches to solving crimes, all of which have proven their value when applied independently, namely: (1) geographic assessment, (2) social networking, (3) crime pattern analysis, and (4) physical description match. The system that integrates these four search techniques, called PerpSearch, takes a description of the crime, including its locations and all other known aspects (e.g., physical characteristics of suspects, vehicles, etc.), and runs it all through the PerpSearch engine components, where they are combined to produce a score for each potential suspect. By using past data on crimes prior to solution and comparing the results against the eventual known perpetrators, the system can be fine tuned and validated. A prototype has been implemented using current Alabama criminal and demographic databases.

IRAS: An Inmates’ Risk Assessment System

  June 1, 2009      Health and Human Services, Law Enforcement
Li Ding, Brandon Dixon, Allen Parrish, International Journal of Computers and Their Applications, Vol. 16, No. 2, June 2009

This research had the goal of improving the classification of offenders into their different levels of risk in order to improve the decision-making process with regard to diversion programs (i.e., alternatives to incarceration). The system is based on an automated assessment of the likelihood of recidivism based on nine weighted attributes. The system can easily be tested by running it on past historical data and then comparing the results with the observation of more recent outcomes. This was pilot tested using data from Madison County, Alabama.