Clustering method and artificial intelligence to classify and project violent crime in Colombia

Abstract

The objective of this research is to propose clusters of violent crimes in Colombia by departments together with a neural network structure for their classification and prognosis. For this purpose, it is based on the analysis of the clustering method, artificial intelligence and the definition of violent crimes. Using a quantitative explanatory method, the data generated by the National Police on crimes between 2018 and 2022 was used. As a result, four clusters of crimes and violence factors that characterize groups of departments were established, which allowed identifying regions with higher and lower impact of criminal acts. A double-layer neural network was then proposed that achieved a classification and prediction capacity of 97.7% for crimes according to their type and impact, which is expected to serve as a tool for crime prevention.