Table 4.

Performance comparison of different classification models, in terms of overall accuracy.

Model	Overall accuracy (%)
Ensemble learning [11]	94.20
BbNNs [18]	94.49 (calculated by us, based on the confusion matrix provided in [18])
End-to-end DNN [15]	94.70 (as the proportion of classes F and Q in the MIT-BIH data set is very small (less than 1%), these two classes have insignificant contribution to the overall performance and so they were not included in the calculation of the overall accuracy presented in [15])
1D-CNN [10]	95.13 (calculated by us, based on the confusion matrix provided in [10])
Improved ResNet-18 (the proposed model)	96.50