I have seen this and this questions, but all of them are about accuracy. I have 5 different binary classifiers on imbalanced datasets (most of the samples are negative). I need to prove that one of the classifiers is better than others. I have the average and std of AUC and Macro-F1 of 5 classifiers on 5 datasets for 10 runs, but absolutely no idea how I should use statistical tests to show the significant of one classifier. I'm familiar with python and R programming, is there any way to compare classifiers with these measurements?