Is it possible to prevent homicides through a warning system in the city of Medellín? How does calling for police help or having a guardian prevent homicides from happening? In which areas of the city cases of homicide are more frequent and where should receive special institucional attention? These issues are explored in the research proposed for the Information System for Security and Social Coexistence (Sistema de Información para la Seguridad y la Convivencia - SISC), a project maintained by the Secretary of Security of Medellín which helps to design and execute public decisions of security and governance. With the goal of creating efficient alerts to prevent homicide through the use of Machine Learning in Medellín, the research provides a better understanding of which methodological approach and variables should be adopted for that.Aligned with Edgelands’ purpose of deepening our knowledge on questions of surveillance and governance in the territories we work in and emerged in the project Decodificando la Seguridad, the authors cross-referenced data provided by the Census as well as open data and satellite images in order to produce a first segmentation of the city which indicates where might be higher or lower attention to the moment when a distress call is received. Headed by Jessica Salazar Vasquez, the work of Hamilton Smith Gómez Osorio, Juliana Restrepo Tobar, Santiago Rivero Ruiz, and Andrés Felipe Rodríguez Cardona combines data with socioeconomic, demographic, and territorial analysis. The problem raised, the methodology adopted, how the data was treated, and the first results were organized in the article ‘Modelos de priorización de atención para la prevención de homicidios’, avaiable here.