A Comparative Analysis of Ensemble Techniques and Hyperparameter Tuning for Enhanced Heart Disease Prediction

The incorporation of machine learning with medical data for prediction in healthcare is in high demand. The accuracy and rapid processing of data play a key role in predictive healthcare data analytics. A small, delayed analytics report with low accuracy, on the other hand, can lead to misdiagnosis, which can lead to severe illness or even death. As the heart serves an important function in the human physiology, diagnosing heart-related issues requires special care. Cardiovascular disease(CVD) encompass a large variety of heart complications including arrhythmia, cardiomyopathy, coronary artery disease, and heart infections. Machine Learning (ML), a branch of artificial intelligence, offers a comprehensive array of algorithms designed for precise and effective predictions. This study uses various algorithms, namely logistic regression, naïve Bayes, random forest, K Nearest neighbor, decision tree, support vector machine and XGBoost with heart data set from kaggle to predict heart disease. To optimize accuracy, GridSearchCV and five-fold cross-validation were employed in the proposed research. Simulations were conducted using a benchmark dataset. The effectiveness of the proposed model was tested using well-known evaluation metrics, such as accuracy, precision, recall, and F-measure. After performing multilevel comparisons with and without fine-tuning of the hyperparameters, the ensemble algorithm LightGBM outperformed the basic and hyperparameter-tuned algorithms with an accuracy of 96.72%.In addition, the web interface "Heart Disease Predictor" was developed to predict the disease based on heart-related attributes collected from the user.