I've been looking into using ROC curves as a evaluation tool of a multi-class classification. The only data I have about this classification is in form of 7-by-7 confusion matrix. Specifically, I've been thinking using Hand and Till's (2001) multi-class AUC, not actually plotting the curves. Is this possible only from a confusion matrix? I am machine learning novice and large part of the Hand and Till's paper is beyond my comprehensions. It is also worth to note that I'd like to implement this in R language.