Utilização de técnicas de inteligência computacional na caracterização de pacientes com doenças cardiovasculares

Abstract

This work encompasses Computational Intelligence (CI) techniques, such as the Vector Support Machine (SVM) and Multiple Linear Regression, in order to classify the 303 patients present in the public database "Heart Disease Database", as cardiac patients or not, based on a series of information designed in periodical examinations carried out in them. In order to reduce or anticipate the diagnosis of cardiopathies, diseases that are at the top of the list of the ones that kill the most around the world, both techniques were chosen for this application, based on previous experiences similar to the one performed in this dissertation, satisfactory performance. Seeking to be a method capable of assisting physicians in the diagnosis of cardiovascular diseases (CVD), this comparison became necessary for the reduction of erroneous diagnoses. From the information collected, we obtained a value of 77% of accuracy, 91% of sensitivity, 69% of specificity and 9% of False Negative in the best simulation for the Support Vector Machine technique, while for the simulations made with selection of variables by Multiple Linear Regression, were obtained 85%, 86%, 84% and 14% respectively, confirming previous studies that show that Computational Intelligence can rather be a helper with the association of simple information such as: age, gender, blood pressure, cholesterol, blood glucose, maximum heart rate achieved, exercise induced angina and ST wave depression, applied to the Support Vector Machine , which in spite of having a slightly lower accuracy, presented a better performance in relation to the False Negative results, thus obtaining a more satisfactory result.