Lags seems not working with tune_test_forecast in MVForecaster

[Coemgenus]

Hello,

Thanks for your wonderful library.

Juste a note, but it seems that in MVForecaster the function tune_test_forecast does not take into account the lags.

When I check the results, everything is similar to use manual_forecast(lags=1)

I’ve tried to use add_ar in every single Forecaster used in MVForecaster but without success

I’m using the last version of the library (0.17.14)

Regards,

[mikekeith52]

Hi,

So to take lags into account in tune_test_forecast(), you would need to add a lags key to your dictionary grid:

mlr_grid = {'lags':[1,2,3,4,5]}
f.ingest_grid(mlr_grid)
f.tune()
f.auto_forecast()

You can see how many lags were selected by then calling

f.best_params

This works the same way for tune_test_forecast(), except now the grids are stored in an external file (MVGrids.py by default) and you can export the lag selections for all models by calling mvf.export('model_summaries').

Let me know if that doesn't answer the question.

[Coemgenus]

Oh i see that’s clear for me, sorry for wasting your time on this one !

Thankyou