Optimization of police response times in Kinshasa through machine learning

Abstract

For some time now, the Democratic Republic of Congo has been facing the problem of juvenile insecurity. This article examined more than 1,042 cases of security incidents in the city of Kinshasa. Most of these incidents are not isolated events but part of broader recurring patterns. This study demonstrates that integrating technological tools with existing law enforcement methods can contribute to the development of a more intelligent, proactive, and efficient information system, thereby strengthening public security in a sustainable manner. The results obtained after data analysis and model testing confirm the hypothesis by demonstrating that the use of predictive tools, in particular logistic regression (GLM) and the generalized additive model (GAM), make it possible to predict the probability of a delayed response time based on the contextual characteristics of the incident.   Keywords: Predictive algorithm, Juvenile delinquency, Predictive policing, Kuluna in Kinshasa, Urban crime; Machine learning, youth violence