Methodology for machine learning models (classification or regression)

Question

I'm wondering if there are any well-justified end-to-end methodologies for developing machine learning models (classification or regression). At work, there is a guy who developed a methodology for classification and I'm wondering if there is some paper or "common methodology" when developing a machine learning model.

When I go to google, I can find stuff like "how to treat outliers", or how-to-do certain steps when building a model. But I can't find an end-to-end approach. Is there any paper where I could look for?

Answer 1

The way I think about it, all ML models are some combination of a) an assumption about the characteristics of the data and b) some level of approximation to an ideal decision/regression function. So the process is to determine if the model represents the salient features of the data and assessing if any approximations it involves are acceptable for the applied problem.

I've found that The Elements of Statistical Learning by Trevor Hastie, Robert Tibshirani, and Jerome Friedman provides a good overview of basic ML techniques in a way that allows the reader to get at the bigger picture.