You're comparing several candidate models for a supervised learning task and need to pick one for production. The team wants a clear rationale for why one model is better than the others, not just a higher validation score.
How would you explain the trade-offs behind your model selection?
