Machine Learning in Cardiovascular Disease Detection: An Experimental Analysis of Techniques

Abstract

Introduction: Cardiovascular diseases (CVD) are the leading cause of mortality worldwide. Early detection is essential for implementing preventive strategies to mitigate serious complications and reduce the mortality rate. In this context, machine learning techniques have become a key tool for developing effective predictive models in the health field.Objective: To improve accuracy in identifying patients at risk of CVD by implementing the wrapper method for feature selection in combination with unsupervised learning algorithms.Methods: Based on the “Cleveland Heart Disease Data Set” dataset from the Machine Learning repository of the ICU KDD. Information Gain and Chi-Square feature selection techniques were applied to identify the most relevant variables in the classification process. Subsequently, several models were trained, including C4.5, Random Forest, SOM, and GHSOM neural networks, and Naive Bayes Tree, to automatically classify the probability of presenting a cardiovascular risk condition.Results: The experimental results show that the Random Forest model, combined with 10-fold cross-validation and the Information Gain technique, achieved the best performance, with a precision of 85.70% and an accuracy of 87.10%. Conclusions: The results of the simulations indicate that the combination of the Information Gain feature selection method with the Random Forest classifier offers the best performance in the identification of cardiovascular diseases, reaching an accuracy that is accepted as optimal compared to the reviewed literature.