Figure 3
The area under receiver operating characteristic (AUROC) curves for the statistical model (a), the XGBOOST model (b) and the Deep learning transformer (c). The curves indicate the probability of the model predictions being correct at various thresholds. The ROC curves and AUC indicate probability of a model attaining a true prediction. The closer the curve is to the top left of the graph, the better. Confusion matrices for the statistical model (d), XGBOOST model (e), and the Deep learning transformer (f). The matrices indicate the normalized count/overall % prediction accuracy in the middle of each tile. The bottom and side values indicate the percentage of correct classifications of the target (High, Medium, and Low) in the column and row respectively.
