Nice one boss,

I enjoyed the explainability of this article, the predicted 11 mins read time seemed like 6 minutes to me because I was engaged all through.

In addition to the effects of class weights balancing which you have brilliantly elucidated , I would like to add that in situations where model explainability to stakeholders is crucial and the effects of false positives is not critical to the usecase being solved for ,simply using the class weights balancing approach may suffice.

Next, I am interested in seeing how these two approaches compare to Synthetic Minority Oversampling Technique (SMOTE) in terms of performance. Go "dataking" ... !