Designing efficient and balanced police patrol districts on an urban street network

In police planning, a territory is often divided into several patrol districts with balanced workloads, in order to repress crime and provide better police service. Conventionally, in this districting problem, there is insufficient consideration of the impacts of street networks. In this study, we propose a street-network police districting problem (SNPDP) that explicitly uses streets as basic underlying units. This model defines the workload as a combination of different attributes and seeks an efficient and balanced design of districts. We also develop an efficient heuristic to generate high-quality districting plans in an acceptable time. The capability of the algorithm is demonstrated in comparison to an exact linear programming solver on simulated datasets. The SNPDP model is successfully implemented and tested in a case study in London, and the generated police districts have different characteristics that are consistent with the crime risk and land use distribution. Besides, we demonstrate that SNPDP is superior to an aggregation grid-based model regarding the solution quality. This model has the potential to generate street-based districts with balanced workloads for other districting problems, such as school districting and health care districting.