Um sistema baseado em aprendizado de máquina para assistência médica no diagnóstico de doenças cardiovasculares

Abstract

In this paper, the early diagnosis of cardiovascular diseases (CVD) through Machine Learning (ML) is investigated, utilizing the database from the University of California, Irvine (UCI). This database was validated in the course of this work with analyses that demonstrated consistency with evidence found in medical literature. Among the ML models employed, the StackingCVClassifier, integrating XGBoost, K-Nearest Neighbors (KNN), and Support Vector Machine (SVM), was highlighted for achieving the highest accuracy, at 92.64%. Furthermore, to prioritize accessibility in areas where infrastructure is limited, a quantitative scale for assessing the difficulty of attribute collection was developed, adapting model training using the database with and without difficult-to-collect attributes. Lastly, to aid clinical implementation, an Application Programming Interface (API) and user interface were developed. Therefore, this paper advances the application of ML in the diagnosis of CVD, proving that the integration of technology, validated data, and accessibility can aid in the early detection of complex health conditions