Modeltime 'updates' Arima specs when running the model on full data

[B81L]

Hi, I am really new to forecasting and modeltime is a really great tool! I just have a question though - When running the training data, and check the accuracy of models, I get the Arima model specs (2,1,1) which seem to be doing a pretty good job on the data, but then when I run the model on the full data in 'updates' to Arima (0,1,1) i use the following code: future_forecast_tbl <- calib_tbl %>% modeltime_refit(data) %>% modeltime_forecast(h = "12 month", actual_data = data)

Is there a way to stop this 'update' and keep the original Arima specs? i look forward to a response. many thanks.

[mdancho84]

When you use auto arima engine, the algorithm is "automatic" meaning it decides which arima model parameters to use.

I don't have an easy way to toggle off the automatic update.

But what you can do is request it as a feature.

In the meantime, you could alternatively create a model just using regular arima_boost engine and specify the arima model parameters. See here.

[B81L]

Many Thanks! that has worked for me - I can tweak the arima boost, based on the auto arima and that gives me some good results!