Detection of Homicide Trends in Colombia Using Machine Learning

The number of violent homicides in Latin America has grown considerably in recent decades, due to the expansion and rise of organized criminal groups in rural and urban areas of the main cities of countries such as Mexico, Colombia and Venezuela. Given their high homicide rate as a consequence of the high crime rate, these countries have been classified among the most violent in the world. According to data reported by the Crime Observatory, the National Police and the Attorney General's Office of Colombia, in 2019 there were 1,032 murders in Bogotá. This data shows a homicide rate of 14.3 per 100,000 inhabitants. From this, it is estimated that between 1960 and 2019, around 226,215 homicides were generated, which is, on average, 3,834 deaths per year. In this work a random forest-based machine learning model is presented, which allows predicting violent homicide (VH) trends in Colombia for the next 5 years. The objective of the model is to serve as an instrument to facilitate decision-making in organizations such as the Prosecutor’s Office and the National Police. The model was evaluated with a dataset obtained from the Criminal, Contraventional and Operational Statistical Information System (SIEDCO in Spanish) of the Prosecutor's Office, which has 2,662,402 records of crimes committed in Colombia from 1960 to 2019.

• Fecha de publicación: 2019
• Institución: Universidad Pedagógica y Tecnológica de Colombia
• Repositorio: RiUPTC: Repositorio Institucional UPTC
• Idioma: spa
• OAI Identifier: oai:repositorio.uptc.edu.co:001/14283
• Acceso en línea: https://revistas.uptc.edu.co/index.php/ingenieria/article/view/11740; https://repositorio.uptc.edu.co/handle/001/14283
• Palabra clave: data mining; homicide; machine learning; random forest; homicidio; machine learning; minería de datos; random forest
• License: http://purl.org/coar/access_right/c_abf302