Updates requested in SARIMA Learner and Predictor components

[nilotpalc]

Hello,
I was trying to use the SARIMA Learner and Predictor nodes.
I was getting the below mentioned errors.

“Unable to import datetime from pandas”
I checked the error message on ChatGPT and understand that the code needs to be revised as below.
from pandas import Timestamp

Also, I am getting an error that knime_io has been deprecated and knime.scripting.io package needs to be used. In the SARIMA Predictor component, only last statement in the code needs to be changed along with the import statement.
knio.output_tables[0] = knio.Table.from_pandas(out)

Also, the Dynamic selection in the SARIMA predictor has the “Output as Integer” selected in the configuration which was resulting in the Python Script failure in the Try-Except loop inside the configuration. Additionally, keeping the Default as selected option is resulting in an error as follows -

ERROR Python Script 3:59:0:376 Execution failed in Try-Catch block: ValueError: In-sample dynamic prediction is not available if memory conservation has been used to avoid storing forecasted or predicted state means or covariances.

It would be great if the default selection for Dynamic can be kept as “Blank” and the Output as Integer option within the boolean configuration node kept as unticked.

I was hoping that the KNIME team can review the Time Series components and update the backend Python scripts for direct use.

Thanks.

–
Best Regards,
nilotpalc

[Corey]

Hi @nilotpalc,
First of all, thanks for the feedback, its very detailed.

Could you verify that the components you’re using are up to date? Either by grabbing fresh ones from the example server under components > time series or by right clicking on a component that is still linked and selecting update link?

Also, is it possible to either share the data set you’re having these issues with or a table spec for it? When I run the components I don’t get these errors at the moment so that might help me reproduce them.

Finally, I’ll check with the Python Dev team to verify what the proper knime.io, scripting, etc, package is to use for the data transfer.

Apologies for the struggles here - we’re hoping a revamp of these components soon though so stay tuned.

[nilotpalc]

Hi @Corey
I have used the updated links from the Components section in the Examples server.
I have added the original components (via the link) and the modified components (post disconnecting the link) and uploaded the workflow in the below link.

SARIMA Forecast Example – nilotpalc

I have uploaded the workflow along with the dataset included for better analysis.

SARIMA Component Failure

SARIMA Predictor Failure

Hope the above info helps.

–
Best Regards,
nilotpalc