
You are working on a supervised learning model and the feature set is large, mixed quality, and partly redundant. Some variables may be noisy, correlated, or only useful in combination with others.
How would you approach feature selection in a machine learning model?