Cyber Detective Links Up Crimes
DePaul University computer scientists Tom Muscarello and Kamal Dahbur have developed a system that compares crime case records with all the files on past criminal offenses using artificial intelligence.
The Classification System for Serial Criminal Patterns (CSSCP) combs through all available case records and assigns distinct numerical values to specifics such as the type of crime, the perpetrator’s gender, and the weapon used, and from them organizes a crime profile; crimes with similar profiles are then sought by a Kohonen neural network. Kohonen is especially adept at uncovering patterns from input data without human assistance.
If a possible connection between two offenses is established, the system compares time and location to determine if the same perpetrators could have traveled between the two crime scenes within the time limit.
Muscarello explains that CSSCP was modeled after conventional crime-solving methodology to a certain degree: Just as several detectives may be assigned to handle individual aspects of a single case, so can the system tap different neural networks to analyze specific angles in an investigation.
Muscarello reports that in laboratory tests using three years’ worth of armed robbery data, CSSCP identified 10 times as many patterns as a team of detectives.
However, the DePaul computer scientist insists that the system is not designed to replace human investigators, but instead serve as a jumping-off point for investigations by finding potentially linked offenses. Muscarello hopes the Chicago police department will agree to run trials of CSSCP.
Abstracted by the National Law Enforcement and Corrections Technology Center(NLECTC) from the New Scientist (12/01/04); Graham-Rowe, Duncan.