Dynamic Analysis for Geographical Profiling of Serial Cases Based on Bayesian-Time Series

Abstract

The analysis of spatial information has long been considered valuable for police agency within the criminal investigative process. This is especially true for serial crime cases where criminologists and psychologist apply geographical profiling to model criminal mobility distribution and behavior patterns in order to estimate a criminal’s likely residence. In recent years the availability of advanced computational mathematic tools ensured us to establish some mathematical models to replace traditional empirical method. However, as a new technology, current geographical profiling models are still fundamental and impractical. In this article, based on existing frameworks, we establish three new methodologies, namely, Bayesian—Factor analysis model, Time series analysis model and GIS(Geographic Information System)—Decay model, to study geographical profiling problems. Then, we test and compare their accuracy, efficiency, sensitivity and robust according to 11 historical serial crime samples and Monte Carlo simulations. Finally, we discuss the advantage and disadvantage of each model and provide executive guidelines about how to synthetically apply these models for real cases.