Comparación de modelos logísticos y árboles de decisión para identificar y predecir factores asociados a la desnutrición crónica infantil basados en la Encuesta Nacional de Salud y Nutrición (ENSANUT) 2018-2019

Abstract

The present qualification work compares two classification models: logistic regression and decision trees in order to establish the best model that determines the factors significantly associated with chronic malnutrition in children under five years of age and makes predictions on this variable. The necessary information was obtained from the repository of the National Institute of Statistics and Censuses (INEC) through the National Health and Nutrition Survey (ENSANUT 2018). The study considered a sample of 11,231 children under five years of age, of which 70% of the sample was used for model training and 30% for validation. The performance of the proposed models was measured by two goodness-of-fit methods: Error rate and ROC curve based on the predictive power of the obtained models. The following results were extracted from the statistical analysis: the logistic regression model was the best for having greater predictive power with an AUC = 62.19%, and a lower error rate of 35%, in addition to the factors associated with malnutrition.