For multi-class classification, the macroaverage is the average of statistical values (prec, recc, etc.) after they have been computed for each seperate class. The microaverage is the combination of a confusion matrix BEFORE statistical values are computed.