Mean of ROC-AUC's aggregated by label.
Build:
python3 py_module_build.py build_ext --inplace
Add to PATH:
import sys
sys.path.append('/Users/tyamgin/Projects/mroc')OR
open ~/.bash_profile and add
export PYTHONPATH=$PYTHONPATH:/Users/tyamgin/Projects/mroc
where /Users/tyamgin/Projects/mroc - path to your module
Usage:
from sklearn.metrics import roc_auc_score
import mroc
label1 = [4, 4, 4, 4]
label2 = [7, 7, 7, 7, 7, 7, 7]
actual1 = [0, 1, 0, 1]
actual2 = [0, 1, 1, 0, 0, 1, 1]
pred1 = [0, 0.4, 0.1, 0.5]
pred2 = [0, 0.1, 0.2, 0.05, 0.06, 0.2, 0]
print((roc_auc_score(actual1, pred1) + roc_auc_score(actual2, pred2)) / 2) # 0.8958333333333334
print(mroc.mean_roc_auc(label1 + label2, actual1 + actual2, pred1 + pred2)) # 0.8958333333333333