Análisis estadístico para la predicción de la probabilidad de encontrar adultos mayores desaparecidos mediante la aplicación de modelos de aprendizaje automático

Abstract

The disappearance of persons is an enigmatic and frequent phenomenon that can have negative consequences for the disappeared person, their family and society in general. Age-related cognitive changes and increased vulnerability to dementia may increase the propensity of older adults to fade. The present study sought to identify individual and environmental factors that could predict whether an older adult reported missing would be found. Supervised machine learning models based on the open data of cases of disappeared persons from Colombia between 1930 and June 2021 (n = 7,855) were used. Classification algorithms were trained to predict whether a missing older adult would eventually be found. The best performing classification models in the test data were: Gradient Boosting Classifier and Light Gradient Boosting Machine, which showed, respectively, 10% and 9% more AUC than a reference model based on the average of elapsed time. since the disappearance. The features that most contributed to the classification were: the time elapsed since the disappearance, the place where it occurred, the age and sex of the disappeared person. The present results shed light on the social phenomenon of missing persons and lay the groundwork for the application of machine learning models in cases of missing older persons.