Explainable Prediction of Cardiac Arrhythmia Using Machine Learning

Cardiac arrhythmia is characterized by abnormal electrical activity of the heart, which can be seen on an electrocardiogram (ECG). Machine learning approaches have shown a lot of promise in terms of assisting in illness diagnosis. In this paper, we propose a method for ensemble classification by combining Heart Rate Variability (HRV), age, sex features and Deep Neural Network (DNN) classification probability features, and achieved the average F1 score of 0.818. It also calculated the importance of features in eXtreme Gradient Boosting (XGBoost) classification decision, and proved the influence of auxiliary inputs (such as HRV, age and sex). This paper shows that explainable machine learning is a promising application in the prediction of arrhythmia.