O efeito da divisão de dados na classificação de arritmias usando redes neurais convolucionais

In this work, we study the effect of dividing the electrocardiogram signal records of different patients between the training and test sets for automatic diagnosis of cardiac arrhythmias. We use convolutional neural network to classify the heartbeats of the signal. The results indicate that using the heartbeats of the same patient in both sets leads to better classification metrics, but is not clinically realistic.