In recent years, the number of people dying from cardiovascular diseases is increasing, and it is a disease that cannot be ignored. Cardiovascular disease has no particularly obvious symptoms before onset and is often overlooked. Clinically, doctors can check whether a patient has cardiovascular disease through a sports ECG. The exercise ECG is an electrocardiogram and blood pressure and heart rate that allow the patient to walk on the treadmill and record their movement status. In the literature, the accuracy of doctors using cardiovascular electrocardiography to diagnose cardiovascular disease falls to about 70-75%. In this paper, we study the use of various machine learning classification algorithms for the analysis of exercise ECG data. The experimental data is data from 100 patients with suspected and known coronary artery disease who underwent exercise electrocardiography after a doctor's diagnosis. We use five machine learning classification algorithms: K-Nearest Neighbor (KNN), Decision Tree (DT), Random Forest(RF), Support Vector Machine(SVM) and eXtreme Gradient Boosting (XGBboost). As a result of the experiment, the XGBboost classifier has the highest performance in all classification models and can achieve an accuracy rate of 84%. The results show that machine learning methods can be used to assist doctors in improving the accuracy of cardiovascular disease diagnosis.