Abstact:

This paper addresses police resource allocation across multiple locations, aiming to minimize the overall cost of potential crimes. Unlike previous literature focused on reactive police tasks, we propose a proactive approach that emphasizes crime prevention through deterrence. To account for the deterrence effect of police resources on crime, we employ the multinomial logit model to calibrate the distribution of crime locations. Our model sheds light on two facets of the deterrence effect in proactive policing—crime control diffusion and crime displacement—relevant to modern crime patterns from both criminology and economics perspectives. We establish the NP-hardness of our problem and provide mixed-integer linear/conic reformulations solvable with standard optimization software. Additionally, we extend our results to a dynamic model over multiple time periods. Finally, we showcase the efficacy of our model through a data-driven case study on the allocation of surveillance cameras in New York City.