Fig. 1
From: Predicting self-perceived general health status using machine learning: an external exposome study
Receiver operator characteristic curves (ROC) for random forest models predicting self-perceived general health. Legend: The left panel a displays ROC curves of models based on the entire 2012 and 2016 datasets, whereas the right panel b shows the ROC curves of model based only on three different types of characteristics (personal, neighborhood and environmental) within the 2016 dataset. Displayed curves are based on the model with the median area under the curve (AUC) value out of the fivefold cross-validated models. The dashed line indicates a ROC with an AUC of 0.5