How will you prevent overfitting when creating a statistical model ?

Answer 1

Given n points, a polynomial of order n-1 will fit them perfectly, but there is only one polynomial with that order that will do the job. So if that polynomial goes to almost infinity between two close points, that’s what you get. That’s obviously overfitting.

To avoid that, don’t use fitting equations with too many parameters, because with, say, “4 parameters you can already fit an elephant”. Somebody said that.

Use fitting models that reflect the physical behavior of your system.

Answer 2

lool I liked the quote hh .. thank u that was so helpful 🤓