Fig. 2From: Comparison of correctly and incorrectly classified patients for in-hospital mortality prediction in the intensive care unitWorkflow. The dataset is split into training and test sets before the ML models are developed using the training set. All models provide an ‘Alive’ (A)/‘Dead’ (D) prediction for all patients in the test set. The procedure of splitting the dataset, training the ML models and making a prediction of all the patients in the test set is then repeated 100 times. All results for each patient are combined and the patients are placed in one of the final groups (TN, TP, FN, FP, or MIXED) depending on the results. This decision fusion is shown in Fig. 3Back to article page