Whenever you're ready, walk me through how you'd approach this.
How do you evaluate the performance of a machine learning model?
Explain a practical evaluation process from defining the task and selecting an appropriate metric through validation, error analysis, and final recommendation. Address how the approach differs for classification and regression, and mention how you would avoid misleading results from data leakage or an unrepresentative test set.