This is a pre-trained deep neural network (DNN) model for risk prediction of atrial fibrillation (AF) from row ECG samples. From ECG data, the model can predict whether a patient is with AF condition, whether a patient is without the condition and without the risk or whether a patient is at risk of developing AF in the near future (within 7 years). The output from the DNN model can then be applied on a survival model for further risk analysis. The companion code can be found in: https://github.com/mygithth27/af-risk-prediction-by-ecg-dnn