Classification of Heart Disease from Clinical Data Using Random Forest and Neural Networks

Heart disease remains the leading cause of death globally. Conventional diagnostic methods are often costly, time-consuming, and invasive. This study proposes a machine learning-based approach for early heart disease detection using Random Forest (RF) and Neural Network (NN) models applied to clinical data. Our methodology integrates statistical hypothesis testing to validate the significance of clinical features prior to model training, which is often overlooked in similar studies. The RF and NN models achieved classification accuracies of 80% and 83%, respectively. Feature importance analysis revealed that ST segment depression and chest pain type are the most influential predictors for classifying and predicting heart diseases from clinical data. The results demonstrate that even with a relatively small dataset, machine learning models can provide competitive and interpretable predictions suitable for real-world clinical decision support, particularly in resource-constrained environments.