I have a multi label – multi class classifier that aims to predict the top 3 selling products out of 11 possible for a given day.
Using scikit learn's OneVSRest with XgBoost as an estimator, the model gets a hamming loss of 0.25.
Im not familiar with HL, I have mainly done binary classification with roc_auc in the past.
Is this an okay score and how can I describe the effectiveness of the model?
does it mean that the model predicts 0,25 * 11 = 2,75 labels wrong on average?
Best Answer
this slide shows a good example, HL=4/(5*4)=0.2
more information refer to https://users.ics.aalto.fi/jesse/talks/Multilabel-Part01.pdf