Optimizing confusion matrix calculations with binary class prediction probabilities

Question

relative Python newbie here... Learning how to construct ROC curves by hand. I have a dataset with a million records and one column that\'s the actual class (0,1) (column is cal