Uso da Inteligência Artificial Explicável aplicada à predição de doenças cardíacas

Abstract

Abstract. The heart is the largest organ in the circulatory system and is essential for pumping blood rich in oxygen and nutrients to all cells in the human body. Problems that occur in the functioning of the heart, such as palpitations, chest pain and shortness of breath, are symptoms that, when severe, are indications of the presence of heart disease. Every year, cardiovascular diseases are among the diseases that kill the most people in the world and many of these deaths could be avoided with effective measures to prevent the disease. Currently, the use of Data Mining has been used to create systems that aim to assist in the diagnosis and prevention of heart disease. Therefore, in this article, Data Mining will be used combined with Explainable Artificial Intelligence to analyze which are the most important characteristics that a model considers when predicting heart disease. Random Forest and XGBoost were the classifiers that obtained the best results, presenting accuracies and recalls above 85%. The results suggest that the use of SHapley Additive exPlanations as an Explainable Artificial Intelligence model was effective in pointing out that the features ST segment slope of the exercise peak and Oldpeak, for example, had a strong influence on the classifiers when predicting the presence of heart disease. Therefore, the use of Explainable Artificial Intelligence can be an important resource to be used in the disease classification process.