Using an Enhanced LightGBM Model to Predict Coronary Heart Disease: Performance Evaluation and Comparison

Abstract

Coronary heart disease (CHD) is a critical cardiac problem that offers a serious health risk and sadly doesn't have a full fix. Detecting coronary artery disease correctly and at an early stage is important for giving effective care to patients. Early identification allows for quick treatments and better patient results. The suggested "HY_OptGBM" model focuses on utilizing an improved LightGBM classifier for predicting CHD. LightGBM is a strong gradient boosting system known for its speed and accuracy in predictive models. The LightGBM algorithm is improved by changing its hyperparameters and improving the loss function. This technique enhances model training accuracy and efficiency. The Framingham Heart Institute's coronary heart disease data helps evaluate the model. By utilizing this data, the model shines in predicting CHD, allowing early diagnosis and possibly leading to reduced treatment costs by treating the disease at its early stages. And also presents a Voting Classifier (RF + AdaBoost) with an amazing 99% accuracy, improving the discovery of Coronary Heart Disease (CHD). This ensemble model, mixing Random Forest and AdaBoost shows stability in distinguishing patterns relating to CHD. To ensure practical usefulness, a user-friendly Flask framework with SQLite integration is integrated, easing signup and signin steps for user tests. This simplified interface improves usability, making the machine learning methods more useful and user-friendly for various parties involved in CHD diagnosis.