Application Analysis of Machine Learning Models Integrating Multi-Source Clinical Data in the Early Prediction of Heart Disease

Since heart disease is one of the main causes of death globally, it makes sense to adopt cutting-edge techniques to increase the prediction accuracy of heart illness, which can lower mortality and save expenses. This research uses data preprocessing, data mining, machine learning, data visualization to generate and evaluate several prediction models, obtain that Random Forest (RF) has highest accuracy but lower ROC-AUC. Although SVM has lower accuracy than RF Model, it has the highest ROC-AUC. Therefore, SVM is the best model for medical scenarios. find the top 5 main factors influencing heart disease are ST_Slope_Up, Oldpeak, ST_Slope_Flat, ChestPainType_ASY and Cholesterol, using data visualization to show the result, which warning people to pay more attention about these physical indicators and seek medical attention promptly in case of mild chest pain or other symptoms. develop a better lifestyle. The research also reflects that there are still many practical challenges of the application of machine learning in data, model, and actual use. The future research could pay attention to improving data qualities, explore better models and gradually enhance practicality.