How do you choose a model?

You are building machine learning features for a spreadsheet assistant. Explain how you would choose an appropriate model for a new problem. In your answer, discuss: - how you would define the task and business objective, - what properties of the data matter when selecting a model, - what practical constraints matter in production, - how you would establish baselines and evaluation metrics, - and the main strengths, weaknesses, and typical use cases of XGBoost. Also explain when you would prefer XGBoost over simpler linear models, random forests, or neural networks.