Problems in the results from "Multivariate Time Series Analysis: LSTMs & Codeles"s

[afgv96]

Hi everyone!

I downloaded the two workflows a few days ago, the training model and deployment model, from this blog: Multivariate Time Series Analysis: LSTMs & Codeless | KNIME.

I have some issues with the results in the training workflow. It’s supposed to look like this:

But instead, I got this:

The predicted cnt is 0.

The screenshot of my training workflow:

I created my Anaconda environment in the Keras section from Preferences/Python Deep Learning.

• python = 3.6.13
• keras = 2.4.0

If anyone can give me a hand and address what I’m missing, it would be much appreciated.

Best regards,

Adrian

[darspir]

Hi @afgv96 ,
do I understand correctly that you ran the Workflow, which you have downloaded, and got different results? Or did you change something in the Workflow after you have downloaded it? If yes could you, please, provide the modified Workflow and, if possible the data, so I could look into it?
Best,
Daria

[afgv96]

That’s right. I ran the downloaded workflow and I got a different result.
I didn’t change a single node of the workflow, I don’t know whether it’s a configuration problem or something else.

[Kathrin]

Hi @afgv96,

I just executed the workflow on my machine using a conda environment which I created via the KNIME preference page. For me the results look like in the blog post and I was not able to reproduce the issue you are facing.

Can you please try to:

1. Create a new conda environment via the preference page and select it.
2. Download the workflow one more time from the Hub.
3. Execute the workflow.

Cheers
Kathrin