Aplicación de técnicas de aprendizaje máquina para predecir problemas cardíacos en pacientes obesos

Abstract

Abstract: Obesity is one of the main public health problems worldwide, from which a large number of diseases and syndromes are derived, such ascongestiveheart failure. Currently, machine learning techniques have been widely used to predict avariety of diseases. Therefore, this research aims to evaluate machine learning algorithms to predict heart failure disease. We worked with a dataset of 412 medical summaries ofpatients who suffer mainly of obesity and overweight. Thisdataset consists of 342 characteristics such as diseases, medications, treatments, among others. Given the large number of features, feature selection techniques such as Chi Square and Mutual Information were used, as well as the PCA dimensionality reduction technique. Five supervised algorithms were applied, of which Random Forest and Naive Bayes stood out with an accuracy of 0.94 and 0.83, respectively, using the Mutual Information technique.