Hello Team I am testing several models, the AutoArima and now including the Sklearn models with SklearnModel in StatsForecast... It turns out that I have the same problem that I had before, and that is that when I make the forecast and visualize the worst Model is the AutoArima, when I use the Cross validation, it turns out that the best model is AutoArima.

models = [AutoARIMA(season_length=season_length),

SeasonalNaive(season_length=season_length),

SklearnModel(Lasso()),

SklearnModel(Ridge()),

SklearnModel(RandomForestRegressor())

]

# Forecast

preds = sf.forecast(

df = train,

h = 120,

X_df = test, # Exogenous variables

prediction_intervals = ConformalIntervals(n_windows = 5, h = 120),

level = [95],

)

# Cross Validation

sf.cross_validation(df = train, h = 120, n_windows = 5)

Hey. Does the same also happen if you remove the exogenous features?

The same problem still occurs, when using cross validation, model selection selects the worst model as the best, and this happens for

Mlforecast

and

statsforecast

, for univariate model and for multivariate

How are you choosing the best model in cross validation?