Predicting Heart Disease with Advanced Machine Learning Techniques

Abstract

Heart disease remains a leading cause of mortality worldwide, highlighting the need for effective predictive tools. However, many studies struggle to report their results effectively. Objective: This study aims to improve the accuracy and efficiency of heart disease diagnosis by employing machine learning and deep learning techniques. It also underscores the importance of raising awareness and implementing preventive measures to reduce heart disease mortality. We utilized the Random Forest algorithm, a supervised classification technique that builds multiple decision trees, to predict heart disease. This algorithm was selected for its capacity to minimize overfitting, handle both numerical and categorical data, manage outliers, and provide insights into feature importance. Hyperparameter tuning was performed using random search to enhance model performance. The effectiveness of the model was assessed with metrics such as precision, F1-score, AUC-ROC, and recall, demonstrating significant improvements in prediction accuracy and reliability.  The Random Forest model achieved an accuracy of 96.7% on the test set. Performance metrics, including precision, recall, and F1-score, were approximately 0.97. The model’s AUC-ROC was 0.92, indicating excellent classification ability.  This study presents a robust framework for heart disease prediction through machine learning, delivering highly accurate results. The Random Forest model’s ability to process diverse data types and its feature importance insights make it well-suited for medical data analysis. Future research should explore additional algorithms and larger datasets to further enhance predictive capabilities.