This study aimed to optimize the homicide predictive classification (HPC) algorithm used by the Military Police of Minas Gerais to increase its predictive accuracy, evaluate the effectiveness of the results over time, and consider its potential for future expansions.
Supervised learning techniques were applied, using data extracted from the police reports system of Minas Gerais state, Brazil, covering the period from January 2012 to June 2023. The data were divided into training and test sets, allowing the construction and validation of predictive models.
The study demonstrates that optimizing the HPC algorithm enhances predictive accuracy, particularly in identifying high-risk individuals. HPC Test 6 outperformed previous models, achieving 2,405 correct homicide predictions within a 66-month horizon.
This study’s limitations include reliance on police reports, which may contain incomplete or underreported data, potentially affecting prediction accuracy. The analysis is also restricted to police incident data from Minas Gerais, Brazil, limiting generalizability. Future research should integrate additional datasets, such as the Integrated Prison Management System (SIGPRI) and Civil Registry, to enhance precision.
The optimized HPC algorithm provides law enforcement with a data-driven tool to enhance crime prevention strategies. By accurately identifying high-risk individuals, the model enables more efficient resource allocation, targeted interventions and proactive policing. Its application supports focused deterrence approaches, improving public safety efforts.
The implementation of predictive policing through the optimized HPC algorithm has significant social implications. By enabling proactive crime prevention, it enhances public safety and fosters community trust in law enforcement. The model supports focused deterrence strategies, potentially reducing homicide rates and violent crime. Expanding predictive analytics to other crime types could further benefit society, promoting a more strategic and effective approach to public security while balancing technological advancements with civil rights protections.
This study contributes to predictive policing by optimizing the HPC algorithm, demonstrating its effectiveness in long-term crime forecasting. Unlike previous models, HPC Test 6 enhances predictive accuracy, enabling proactive interventions. The research integrates criminological theories with machine learning, offering a novel approach to homicide prevention. Its findings provide valuable insights for law enforcement, emphasizing data-driven decision-making. The study’s methodological advancements highlight the potential for expansion to other crime types, reinforcing its originality. By addressing predictive policing challenges, this research enhances public security strategies and contributes to the broader discourse on technology-driven crime prevention.
