Enhancing Heart Disease Prediction: A Comparative Study of Hyperparameter Optimization Techniques with Boosting Algorithms

Cardiovascular disease (CVD) continues to be the foremost worldwide contributor to mortality, leading to millions of deaths each year. Machine learning methods provide a powerful means of prediction. By employing patient data encompassing factors such as age, blood pressure, and cholesterol levels, machine learning (ML) models empower precise predictions, facilitating early detection and focused intervention. This study’s main goal is to investigate how machine learning algorithms can be used to predict heart illness, utilizing a data set with 1000 instances and 12 features from an Indian multi-specialty hospital obtained from Mendeley. The research primarily utilizes four boosting-based machine learning approaches, including Light Gradient Boosting (LightGBM), Extreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), and Categorical Boosting (CatBoost), to predict cardiac disease. To enhance the performance of each algorithm and identify the best optimization strategy, we explored three methods: grid search, random search, and Bayesian optimization. Precision, accuracy, F1 score, and recall were used to evaluate the models. Significantly, the use of CatBoost combined with random search hyperparameter optimization yielded superior results compared to the alternative models, attaining a noteworthy accuracy rate of 99.00%. Our findings suggest that the suggested approach might be a reliable way of predicting cardiac disease in clinical settings.