Your question is Imbalanced Spam Detection Metrics. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How do you handle highly imbalanced datasets when training a spam detection model, and which evaluation metrics would you prioritize?