Regresar

Evaluation of Efficiency in Modeling Mental Illnesses. Case Study: Ecuador

Abstract:

Mental illnesses represent a globally prevalent disease whose mitigation requires the development of efficient mathematical models through the use of machine learning algorithms. Linear regression models, random forest models, and neural networks are the appropriate techniques to explain or predict mental disorders. In this context, descriptive, exploratory, and non-experimental methods were applied to analyze macroeconomic variables in Ecuador (prevalence of mental disorders, GDP, unemployment rate, population, and years of registration). The advantage of estimating regression models has been the ease with which they can interpret their parameters and predict continuous values, with a mathematical structure that can be complex and adaptable to cause overfitting in parameter estimation. Regarding the random forest model, it showed the best results, with an accuracy of 87.88% and a mean absolute error of 0.46. This indicates its versatility in explaining the causes inherent in developing mental disorders and achieving a prediction of the prevalence of these diseases in the Ecuadorian population. The structure that defines artificial neural networks can learn complex patterns in data and can be used to predict both continuous and categorical values. These findings are necessary to predict future values for the variable prevalence of mental disorders and conclude that by applying machine learning algorithms in the construction of models, greater precision is achieved in the identification of the factors incident to the level of risk for the development of mental health disorders in a global environment.