This article analyzes the incorporation of predictive statistical models to support the implementation of predictive policing approaches in crime prevention within large urban centers. It discusses documented experiences in recent specialized literature regarding the development and refinement of predictive models based on machine learning, their application within the predictive policing framework, and their limitations. The study critically examines the effects of machine learning in the broader field of predictive criminology, particularly concerning homicide reduction. It concludes with a discussion of the potential contributions of predictive criminology in the management of public safety within large urban centers.