Resumen

SILVA URRUTIA, Jose Eliud  y  VILLALOBOS, Miguel A.. An empirical analysis of homicides in Mexico through Machine Learning and statistical design of experiments. PSM [online]. 2022, vol.20, n.1, pp.1-27. ISSN 1659-0201.  http://dx.doi.org/10.15517/psm.v20i1.48217.

Homicide is one of the most important mortality causes that has reduced the Mexican life expectancy. That is why the aim of this work is to identify some sociodemographic and economic factors that can help explain homicides in Mexico and measure their impact, assuming the current conditions prevail. To do that, several Machine Learning (ML) methods were evaluated. The C5.0 model is best suited for the data at hand. After fine-tuning the algorithm, we used the estimated model to identify the main factors that explain homicides. Among these factors, eleven were selected that can be influenced by direct changes in domestic public policy, laws and/or regulations. These were used as input in a two-level fractional factorial Statistical Design of Experiments (DOE) to estimate their main effects and possible interactions. Although several of these factors had statistically significant effects on homicide rate, the one that had the biggest and direct impact from a practical perspective, was the Rule of Law Index (RLI). In fact, if we assumed that all states had the median RLI of 0.37, implementing domestic policies and procedures to move them all to the best RLI level could significantly reduce homicide rates.

Palabras clave : mortality; homicides; C5.0; Machine Learning; Statistical Design of Experiments.

        · resumen en Español     · texto en Inglés     · Inglés ( pdf )