Hi again , i have run the below with MLForecast with custom features ml_models = [ LGBMRegressor(random_state=0, n_estimators = 1000, verbosity=-1), XGBRegressor(objective='reg:squarederror', n_estimators=1000, seed=123), CatBoostRegressor(iterations=1000, random_seed=42, silent=True) ] # Forecast using MLForecast for traditional ML models ml_fcst = MLForecast( models=ml_models, freq='D', lags=[1, 7, 14, 28, 90, 180, 360], lag_transforms={ 1: [ExpandingMean()], 7: [RollingMean(window_size=7)], 28: [RollingMean(window_size=28), ExponentiallyWeightedMean(alpha=0.5)], 180: [RollingMean(window_size=28), ExponentiallyWeightedMean(alpha=0.3)], 360: [RollingMean(window_size=28), ExponentiallyWeightedMean(alpha=0.3)] }, date_features=['year','quarter','month','dayofweek', hf.is_holiday,hf.is_easter_related], target_transforms=[Differences([1])] ) Is it possible to run something similar for features with Auto Models like AutoCatboost etc? it is not very clear how can i do this?`First of all can i pass these to the automodels? can we have optuna to perform feature selection out of these?

Do you mean the lags, lag_transforms, etc?

Hi , yes. lag, transforms, target transforms

That's explained here, is there something that isn't clear there?

Hi, somehow i missed this in the doc . When I apply the feature tuning does hyperparameter optimization take place?

The optimization tries model parameters as well as features and transformations. There are defaults for these but you can override them as well