What is ROC-AUC and how is it used in predictive analytics? How does the ROC curve represent the performance of a classification model? What does the AUC value indicate about model accuracy? How is ROC-AUC different from other evaluation metrics like precision and recall? In what situations is ROC-AUC the most appropriate metric to use?