Fig. 3

Optimization and validation of RED_SNN model using 5-fold cross validation. a ROC curves to evaluate the prediction accuracy of the RED_SNN model. The model was serially trained to learn patient’s survival status within a 10-month time interval, until 62 months from the initial observation. ROC curves predicting survival at each time interval were evaluated with validation sets. b The mean AUC’s of the survival prediction at each time point. The average AUC of 5-fold cross validation was 0.84