Personalized Heart Disease Prediction Using Data-Driven Machine Learning Approaches

Cardiovascular diseases (CVDs) are persistently projected as one of the current major health concerns across the globe, thereby emphasizing the importance of an accurate and personalized prediction model. The typical predictive models currently used for health-related diagnostics are mostly based on general models and clinical screening, and in some cases, they are incapable of examining nonlinear interconnections among specific patient risk factors. To address this shortcoming, we propose a machine learning model for personalized heart disease prediction. Multiple supervised machine learning models, namely Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naïve Bayes (NB), and Random Forest (RF), are developed and compared by using the popular UCI Heart Disease dataset. Extensive preprocessing and normalization techniques are used in this study to improve prediction accuracy. Our proposed models show the performance evaluation processes using ROC-AUC, learning curves, and calibration analyses, which justify the accuracy, applicability, and interpretability of the models. The results of this study show that Random Forest's cardiovascular classification, with an accuracy of 98.01%, a Precision of 97.90%, a Recall of 97.99%, and an F1-score of 98.00%, outperformed all other machine learning models.