Abstract：Electrocardiogram(ECG) plays an irreplaceable role in the diagnosis of cardiovascular diseases, and machine learning shows a unique advantage in the automation of ECG diagnosis. A total of 21 837 ECG records in the public ECG database PTB-XL were used. The ECG data were preprocessed by deleting missing values, clipping and removing baseline before extracting their relevant features by principal component analysis. Classification research was carried out using K-nearest neighbor, random forest, Logistic regression and support vector machine(SVM) algorithms, and the performances of the four algorithms were compared and analyzed according to the research results. The results show that the SVM algorithm is obviously superior to the other three classification algorithms in aspects of model evaluation indicators including accuracy rate, recall rate, precision rate, and area under curve(AUC).
