FIRST: FRAMEWORK TO INTEGRATE RELATIONSHIP SEARCH TOOLS

Li Ding, Dana Steil, Brandon Dixon, Nicholas A. Kraft, David B. Brown, and Allen Parrish

Keywords

Data mining, decision tree, crime detection, geographic profiling,crime pattern

Abstract

Information technologies such as data mining and social network analysis have been widely used in law enforcement to solve crimes. Recent researches indicate that geographic profiling also plays an important role in the investigation of crimes. In this paper, we propose an integrated crime detection framework named FIRST that will take given descriptions of a crime, including its location, type and the physical descriptions of suspects (personal characteristics or vehicles) as input. The system will process these inputs through four integrated components: geographic profiling, social network analysis, physical matching and crime patterns. Essentially, geographic profiling determines “where the suspects are, while other components determine “who the suspects are. We then process the results using a score engine which is optimized by a genetic algorithm to give investigators a ranked list of individuals. A hundred real burglary and robbery cases in the city of Anniston, Alabama are used in our experiments to evaluate FIRST. Our experimental results show that the median ranking of arrestees remains at the top 2% of the list.

Important Links:



Go Back