Skip to Main Content
This paper discusses the use of cognitive psychology and data mining to construct a predictive model of crime. This model predicts from the location, time, and daily mean temperature whether the theft was of a bicycle, a firearm, or a purse. It also discovers the features that were salient to the choice of a target for these three crimes. The model was constructed for Richmond, Virginia. In this analysis association rules were used to construct an independent set of crimes. Then a classification and regression tree methodology was used to create a classification tree. This tree was used in a predictive model that, given the location of the crime, the mean air temperature on that day, and the time oft he crime, predicted the type of item stolen. The resulting model predicted the object of the theft with accuracy significantly above that of a random draw.