Optimised Computational Approaches for Heart Disease Prediction Using Dual Stage Machine Learning and Deep Learning

Heart disease continues to be a major global health challenge, ranking among the leading causes of death worldwide, making early diagnosis very crucial for improving patient outcomes and reducing healthcare costs. This study presents the development of a machine learning model aimed at predicting heart disease using health indicators and this data is sourced from the 2015 BRFSS (Behavioural Risk Factor Surveillance System). A stacked ensemble learning approach was utilized, combining multiple base models—Random Forest, Gradient Boosting, and XGBoost—integrated with a Logistic Regression meta-model. The dataset was preprocessed through standardization and one-hot encoding to prepare the features for model training. The model was trained using 80% of the dataset, while the remaining 20% was used for testing, achieving an accuracy of highest possible. The model showed strong performance in predicting individuals without heart disease, though recall for detecting positive heart disease cases was lower, indicating a need for further refinement. This study highlights the effectiveness of stacking in predictive healthcare models and emphasizes the importance of addressing class imbalances to improve recall in identifying heart disease. Future improvements may focus on fine-tuning the model to enhance its sensitivity and applicability in real-world medical diagnostics.