Hello everyone,
I did classification on the GEE platform with Sentinel. Then I collected ground sample points for kappa accuracy analysis. But I also want to create accuracy assessment with ROC Curves and AUC. I have excell tables with predicted and actual class from GEE. I generated ROC Curve plots and AUC from these values using the One vs All algorithm in R. I did this for each land cover class.
My question is: What should I do to find an overall AUC? What is the difference between the overall accuracy value I get from the correctly predicted points and the AUC value for the classified year?