CARDIOVASCULAR DISEASE PREDICTION USING RANDOM FOREST MACHINE LEARNING ALGORITHM

Abstract

Every year, cardiovascular disease (CVD) claims the lives of nearly 17 million people worldwide. Predicting heart disease early and accurately can help delay therapies and improve results. Patient data analysis machine learning techniques have shown promise for better predictive capabilities than conventional methods; however, there are still gaps in areas such as algorithm blending, standardization, feature optimization, and model tuning that require strong methodology. By benchmarking against established methods, this study attempts to create a more sophisticated machine learning model with detailed performance and a robust approach for predicting heart disease. Using a clinical dataset that was obtained from an internet repository, an improved random forest (RF) model was created. It was then tested against baseline logistic regression and support vector machine models, Naïve Bayes Classifier, K Nearest Neighbors Classifier, and Decision Tree Classifier. RF hyperparameter tweaking, redundant feature filtering, and systematic data preprocessing were used. Accuracy, precision, recall, F1 score, and ROC analysis were computed as evaluation measures. With F1 score, 1.00 AUC, and 90% accuracy, The RF model demonstrated superior performance compared to the remaining models, which exhibited, AUCs of 0.9, 0.82, and 0.9. On the public dataset, the refined RF model demonstrated exceptional predictive performance, highlighting the promise of a methodical machine learning approach to improve heart disease prediction. The external clinical validation and optimization of various patient populations should be the main areas of attention for future research.