Measuring AUPRC in CatBoost

Question

I want to measure area under the curve of precision-recall curve (AUPRC) in catboost, but the CatBoostClassifier, doesn not have AUPRC as an evaluation metric.Any suggestion that helps me to measure this performance metric will be appreciated.

This is the code I use:

model = CatBoostClassifier(
custom_loss=['Accuracy','Precision','Recall','F1','AUC'],
random_seed=42,
logging_level='Silent')
model.fit(
X_train, Y_train,
eval_set=(X_test, Y_test), plot=True)

score 0 · Answer 1 · answered Oct 16 '20 at 14:26

0

Catboost has MAP metric which stands for Mean Average Precision. Pretty much the same as the AUCPR. This is a good reference article explaining why.

answered Oct 16 '20 at 14:26

Leopoldo Corona

1

Could you summarize what the article says in a sentence or two? – Tatiana Racheva Oct 16 '20 at 17:07
Use references as additional resource and not the main source. If the link is no longer available on the future your answer becomes useless. – Gi1ber7 Oct 16 '20 at 18:26

score 0 · Answer 2 · answered Oct 06 '21 at 09:33

0

To use average precision as metric you can use eval_metric="PRAUC:use_weights=false which has the same meaning with scikit-learn average_precision. This metric is implemented since catboost 0.32.1 version.

answered Oct 06 '21 at 09:33

zonna

46
1
9

Measuring AUPRC in CatBoost

2 Answers2

Linked