Prediction of Cardiovascular Diseases Using Machine Learning Models

Abstract

The study addressed the global problem of cardiovascular diseases, which were one of the leading causes of mortality and morbidity according to the World Health Organisation. Multiple risk factors, both modifiable and non-modifiable, were identified, and the need to implement technologies that would enable early and accurate detection was emphasised. Given this scenario, the use of machine learning algorithms such as Support Vector Machines (SVM) and Artificial Neural Networks (ANN), combined with traditional and alternative kernel functions, was proposed. A comparative approach was developed to validate the hypothesis that under-explored kernel functions could improve predictive performance in terms of accuracy and response time. To this end, models were trained with data extracted from recognised platforms such as Kaggle and UCI, and metrics such as accuracy, recall and F1-score were applied. The models were adjusted with hyperparameter optimisation techniques using random search. The results demonstrated that certain alternative kernel functions offered improvements in the error-time ratio, in some cases outperforming conventional kernels. The research not only contributed methodological advances in the development of predictive models, but also provided a support tool for clinical decision-making, particularly useful in contexts where timely diagnosis is crucial. Finally, the project contributed to strengthening artificial intelligence in public health, promoting well-being through the prevention and proactive management of cardiovascular diseases.