medium10 pts · first answer countsmodel-evaluationstatisticsconfidence-intervals

Error bars on a test metric

How do you get a bootstrap 95% confidence interval for a model's test F1?

A.Resample test rows with replacement, recompute F1, take percentilesB.Retrain the model 95 times on the same data and average F1C.Multiply the single test F1 by 1.96 to get each boundD.Average the F1 on the training set and on the test set

Discussion 0

Join free to post in the discussion.

Error bars on a test metric

Discussion 0

Related questions