Is there any was to plot the AUC and ROC from the confusion matrix?
If not, how should I implement them and, more specifically, where? Before the validation?
Surely this isn’t possible. A confusion matrix provides the values of TP, FP, TN and FN at a given classifier threshold i.e. a probability value, corresponding to a single point on a ROC curve. Conversely, if you already have a ROC curve (and hence AUC), then you can generate a different confusion matrix for any point on the curve.
Can you guide me on how to implement ROC curves? I tried at least 5 ways to implement them with no success.
I have 5 classes.
Well, I can try 
ROC analysis is usually used for binary (2-class) classification so with multiple labels you plot one curve for each class. This is a helpful link to read: here.
Have a read and let me know how you get on.
Anthony.