Comparación de algoritmos de aprendizaje de máquina y técnicas de selección de variables para la predicción de depresión

Abstract

Abstract: Depression is a mental disorder that has affected the quality of life of millions of people, and its early detection has been a challenge for healthcare professionals. Given this problem, this thesis focuses on the use of supervised learning algorithms based on trees in conjunction with different variable selection techniques to improve the accuracy of predictive models for the detection of depression. To this end, data from a nationwide survey of 3,300 people suffering from some degree of depression will be analyzed, including demographic, sychological, and substance use information.