Pattern Recognition and Analysis: Neural Network using Weighted Cross Entropy

Abstract

In order to identify arrhythmia more simply and efficiently, one-dimensional convolutional neural network is used to extract electrocardiogram signal features and a neural network model based on cardiac beat is established, which can identify eight major ECG arrhythmia types. The cross entropy loss function is weighted according with the data distribution to solve the data imbalance problem. Experiments verify the effectiveness of the weighted cross-entropy loss function in ECG recognition, which improves the accuracy of all kinds to more than 99.5% and the sensitivity of all kinds to more than 90%. Compared with other models, this model can improve the performance of ECG arrhythmia recognition.