Your question is Extreme Imbalance in Fraud Detection. 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).
You are training a binary classifier to detect ad fraud, where positive examples are very rare and the cost of missing fraud is high. You need a training and evaluation approach that works when standard accuracy can look strong even if the model is not useful.
How do you handle extreme class imbalance when training models for ad fraud detection?