Simple filtering options with sequential search.

[pinehowe]

The sequential search method for variable selection is excellent, although there are some simple tweaks that could make it easier to use, like adding in a variable correlation filter for forward selection. That would be an easy way to avoid collinearity issues with some tree-based machine learning models like XGBoost & random forest where including collinear variables decreases the interpretability of individual variable effects.

[be-marc]

Thanks for the proposal. I think you could already do this with mlr3filters, mlr3pipelines and maybe an AutoFSelector. Or are you missing something?

[pinehowe]

Thanks for the proposal. I think you could already do this with mlr3filters, mlr3pipelines and maybe an AutoFSelector. Or are you missing something?

That may be the case, but I haven't been able to figure out how to integrate the filter while using fselect(). Perhaps an example in the documentation would be sufficient to help people. Anyway, thanks for the help!