Predictive Analytics in Heart Disease: Leveraging LightGBM for Improved Diagnostic Accuracy

The prevalence of heart disease is significant, making it a prominent health concern, and early prediction and diagnosis are critical. The application of artificial intelligence algorithms in heart disease prediction is highly promising. This study utilizes the LightGBM algorithm for the predictive modelling of a heart disease dataset containing 1025 records and 14 health indicators. Predictive models were effectively built through data preprocessing, feature selection, and model training. The results show that the LightGBM model got 98.54% accuracy in the test set is better than the model of support vector machine (SVM) and Logistic Regression. Age, heart rate, and cholesterol levels are the main factors identified by feature analysis. The Area Under the Curve (AUC) value of the model reaches 0.985, proving its excellent performance. Despite this, the model's generalization ability may be impacted by the dataset's size and representativeness. Future research may take into account the use of larger and more heterogeneous datasets to further verify the effectiveness of the LightGBM algorithm in heart disease prediction. This study shows the effectiveness of the LightGBM algorithm in predicting heart disease, which can provide support for making medical decisions and public health management.