Hi,
My team and I were wondering why it is required that we output a column "was_preterm" where we predict a 0/1 label for each subject? Is this column taken into account in the evaluation? We were under the impression that the evaluation is based on auROC (and auPR in the event of ties), which should just depend on the probabilities we predict. Any clarification that could be provided would be much appreciated!
Best,
Will K
Created by William Kindschuh wkindschuh Hi @wkindschuh,
The class labels are used to calculate accuracy, sensitivity, and specificity and therefore provide insights into model performance. They also may be used in post-challenge analyses.
Tomiko on behalf of the 'Preterm Birth Prediction - Microbiome DREAM Challenge' team