The paper only evaluates the accuracy of the scores given a (quite arbitrary) threshold of 5. However I believe that the spearman rank correlation is a better metric in this case. Even though your model is trained for binary classification (and not regression), wouldn't it make sense to use this metric?
Hello,
The paper only evaluates the accuracy of the scores given a (quite arbitrary) threshold of 5. However I believe that the spearman rank correlation is a better metric in this case. Even though your model is trained for binary classification (and not regression), wouldn't it make sense to use this metric?