Week #10: performance evaluation and validation

Read the three suggested papers and answer:

1. What are the problems related with evaluating models in the context of cross-validation? Discuss about micro-averaging and macro-averaging and their impact in the evaluation of corss-validated models.
2. What is the relationship between ROC and PR curves? Why the second paper argues that ROC produces overly optimistic results when the class is imbalanced?
3. What is the argument of paper #3 when saying that it is OK to use ROC with imbalanced classes?