I'm trying to build an xgboost regressor or a catboost regressor for a task. I have a working linear regression model. I also trained an xgboost regressor model for the task but it was worse than the linear regression model. I am wondering if there is a way to pass the linear regression weights (model parameters) as an initial set of parameters to the xgboost (or catboost) model to ensure performance gain?

E.g. if $w0*x0 + w1*x1 + w2*x2 + w3*x3 = y$ is the linear regression model, is there any way to tell xgboost to start by the same equation (and get better while training)?


1 Answer 1


Answer is NO

Why? weights are hyperparameters of the linear regression model and they are not the same as the ones for xgboost or catboost.

What you can do is combine models (if you really want to use xgboost or catboost additionally) SEE this

NOTE just because model is more powerfull it does not mean that it will beat linear regression on every dataset ;)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.