The original dataset is split into a Risk Evaluation set (20% of samples) and an intermediate dataset (80%). The intermediate dataset is split into the model training dataset (60% of total sample) and a risk determination dataset (20% of total sample). The model training dataset is used to train 60 cross-validation models. The results of these models are used on the risk determination dataset to define the risk levels and score thresholds. Finally, the risk evaluation dataset is used to evaluate and compare the observed prevalence on the different risk levels.