Smart Patrolling Based on Spatial-Temporal Information Using Machine Learning

Abstract

With the aim of improving security in cities and reducing the number of crimes, this research proposes an algorithm that combines artificial intelligence (AI) and machine learning (ML) techniques to generate police patrol routes. Real data on crimes reported in Quito City, Ecuador, during 2017 are used. The algorithm, which consists of four stages, combines spatial and temporal information. First, crimes are grouped around the points with the highest concentration of felonies, and future hotspots are predicted. Then, the probability of crimes committed in any of those areas at a time slot is studied. This information is combined with the spatial way-points to obtain real surveillance routes through a fuzzy decision system, that considers distance and time (computed with the OpenStreetMap API), and probability. Computing time has been analized and routes have been compared with those proposed by an expert. The results prove that using spatial–temporal information allows the design of patrolling routes in an effective way and thus, improves citizen security and decreases spending on police resources.