
You're iterating on a supervised learning model and need to improve its performance without overfitting. You want a structured way to decide whether to change features, model complexity, regularization, or training setup.
How do you optimize the performance of a machine learning model?