Your question is Feature Engineering for Transaction Risk. 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 would you use feature engineering to improve a model for transaction risk?
Discuss how you would transform transaction, customer, merchant, device, and payment-history data into predictive features. Cover leakage prevention, class imbalance, temporal validation, interpretability, and how you would measure whether engineered features improve the model. Provide a practical implementation outline using Python and scikit-learn.