Hi Nixtla team 👋! I'm working with

mlforecast

and exploring ways to automate feature engineering. For feature creation, I think I see paths forward (e.g., adding/excluding custom feature groups as part of optuna trial). However, I'm looking for guidance on automated feature selection within the

mlforecast

pipeline itself. I considered using the

Optuna

integration, but that seems more geared towards overall hyperparameter tuning rather than specifically iterating on feature sets before the main model tuning/training. Is there a recommended way to achieve automated feature selection with

mlforecast

currently? Or, perhaps, could a small module dedicated to feature selection (e.g., based on importance, stepwise methods) be a potential future addition? That would be incredibly helpful! Any advice or pointers would be greatly appreciated! Thanks! 🙏

Most common ML methods do implicit feature selection already (e.g. tree methods like LightGBM). You can always perform feature selection, but the best way of doing that (imho) would be a cross-validation type approach based on Shapley values, what ING did here. This will become computationally heavy very fast, and generally time is better spent on feature engineering and having the MLforecast alg decide on the feature importance internally.

I dont particularly agree with the statement that lgbm does feature selection internally inclusion of certain feature sets can lead to poorer perf. but disregarding that I have for now opted to include it as part of the optuna config and I think it does the job good enough... I will look into the link shared... thank you for your assistance