A novel serial crime prediction model based on Bayesian learning theory

Abstract

How to build affective mathematical models to understand the behaviors of serial crimes is an interesting research field in public security. Several theories have been proposed to handle this problem. In this paper, we introduce a novel serial crime prediction model using Bayesian learning theory. There are many potential factors affecting a serial offender's selection of the next crime site, we mainly studied the factors related to geographic information. For each factor, by using a discrete distance decay function which derives from the classical crime prediction theory “Journey to Crime”, we create a geographic profilewhich is a probability distribution of being the next crime site on given geographical locations. The final prediction is made by combining all geographic profiles weighted by effect functions which can be adjusted adaptively based on Bayesian learning theory. By testing the model on a crime dataset of a serial crime happened in Gansu, China, we can successfully capture the offender's intentions and locate the neighborhood of the next crime scene.