Problem with an unconnected node


I have this problem in my ARIMA time-series. The error is:
"Contains an unconnected node (“component output …*…)”.

The excel reader not deprecated dont change the situation, but change the numbers (in the …*…).

Plese help :cry:

Hi @stefano_c and welcome to the community!

Have you considered updating the components? It looks like you are using an old version: The new version uses a python port to pass the model along: ARIMA Learner – KNIME Hub and ARIMA Predictor – KNIME Hub . You can update either by right clicking the component → Component → Update Link (given that the link to the Hub is correct) or by freshly dragging in the component from the Hub.

It could also be that the error is misleading and maybe it helps to execute the preceding Excel Reader node? Did this workflow run through in the past? It’s a bit weird, this would mean a non-working version was updated to the hub. Could you please share the workflow if above does not work for you?

Best Regards,

Hi Lukas, thanks for the fast reply.

The execute of excel writer dont change, but the different Arima Learner produce a different error.

Look in my workflow
KNIME_project4.knwf (235.6 KB)
Thanks again!

Hi @stefano_c,

unfortunately I can’t use your workflows as I’m missing the MMCO2SET.xlsx - could you share this, please? Best is to store it in the data area of the workflow and link it via relative paths (see e.g. this example: Storing a Data File inside the Workflow Data Area – KNIME Hub) :slight_smile: You can also upload the executed workflow to the hub and share it here.

If the data is too sensitive, could you elaborate on the error what exactly does it say? Is python correctly configured and the statsmodel package installed? If you don’t want to use python, you can also try these nodes KNIME Autoregressive integrated moving average (ARIMA) – KNIME Hub


Sorry, it not a problem share the excel.
MMCO2SET.xlsx (13.9 KB)

These nodes “KNIME Autoregressive integrated moving average (ARIMA) – KNIME Hub” use a different Estimation method, but i not sure. I would see the old process to compare the different nodes with the hope to have the same results.

Hm, the component in your workflow seems to be in a strange state where it has two output nodes - something went wrong there somehow, and its hard to tell where and why. I’d suggest to start over with a freshly dragged in component: Here is a version that hopefully works for you:

(might take a while during the first execution, as a new conda environment will be installed) It looks like the python methods that are used under the hood are sensitive as to which versions to use (statsmodels <=12.1) since conda environment propagation nodes where not around yet when the component was developed. I’ll try to change that - please let me know if the modified workflow already works for you!

Best Regards,

1 Like

Thanks for this workflow, it’s very interresting.
Unfortunately dont work, and do this error

I already create the conda environment, but dont work anyway.
How can I solve this problem? Create a new environment again?

Or i need a Python workflow to input in the Knime workflow?

Ah, my bad: I forgot to reset the Conda Environment propagation node - please reset and execute it yourself or download the modified workflows. This will install the conda environment itself, so that it is exactly the same as I used on my machine.

No problem.
Unfortunately, dont work again.
Same problem: dont find “statsmodel”

How is possible?

1 Like

Oh, I forgot to tell the Python Predictor to use the conda environment - please try again.

If you want, you can adjust this yourself by opening the component (right click → Component → open or Ctrl + double click), configure the Python Predictor nodes like so:

Hopefully this does the trick now? :crossed_fingers:

I’m sorry, i dont understand the process.
When and where you use the node Python Predictor?
I can see the workflow under the dialog?

This is inside of the ARIMA Learner Component which you can open as described above. But I already uploaded the corrected version of the workflow, so you can also just use this one:

1 Like

My bad, i understand.

Hurra! it works! Thank you so much.

Could this Python problem happen again? Is better disinstall and restart Knime, and reinstall the conda environment? This trick can be used in other situation with is necessary “statsmodels” or another conda patch?

Anyway, thank Lukass for your help and for the solution.

Best regards,


Wohoo, that’s great :slight_smile:

In this case it should be solved: This problem happens when you get a workflow/a component from someone else who uses a different conda environment than yourself - so I’d always advice to use the conda env propagation node.

If you want to use the “py39_knime_arrow” env permanently in KNIME you can go to File → preferences → KNIME → Python and set it there:

This way the ARIMA Learner and Predictor nodes work “right off the Hub”. Otherwise I’d advice to use the conda env propagation node.

Have Fun!

1 Like

One more thing for the future: @Corey just adviced to use the SARIMA Learner and Predictor Nodes - they have the conda environment node built in and can handle the specific ARIMA case as well

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.