ECG Signal Classification using Long Short-Term Memory (LSTM) Networks

This paper studies on learning rate impacts the performance of an LSTM model in the context of ECG classification. ECG classification is vital for diagnosing and monitoring diverse cardiovascular conditions including atrial fibrillation. To investigate the effect of the learning rate, the LSTM model is trained with various value of learning rates, and subsequently, the classification performance is evaluated. According to the findings, a learning rate of 0.006 yields the highest accuracy compared to other values. By employing the learning rate, the model undergoes training using feature-extracted data and achieves an accuracy of 93%. The experimental shows the learning rate selection is important for achieving optimal classification accuracy and robustness in ECG classification tasks.