Export Gradient Boosted Trees Predictor (Regression) to Excel

Hi Community,

I am trying to find out, if I can export a trained Gradient Boosted Trees Predictor (Regression) to Excel in order to calculate in Excel with Input Data the prediction.
Is there a (direct) way to export the model to excel?

Thanks for your comments with regard to my question

Bernd

(P.S. - I am a new member if KNIME Plattform)

No.

And it never will be because how should that work?

@Bernd welcome to the KNIME community.

Is there a specific reason why you would not want to use KNIME to generate the predictions?

About your question. There is (was?) Zementis which promised to bring PMML models to a lot of platforms, including Excel. Assuming you could convert you model into PMML.

On the KNIME site there are still presentations from a few years back about the platform (before it was aquired).

There is still a support site mentioning some Excel plugins - but I am not sure it is well maintained
https://support.zementis.com/hc/en-us

Maybe you just stick with the KNIME platform itself.

2 Likes

@mlauber71:

Many thanks for your answer. I will contact Zementis and check it out.
But openly spoken I hoped to find an easier way to export the model (and not via PMML and additional software toolsā€¦)

Perhaps I should explain the task a little bit deeper, which I have:
I tested a gradient boosting machine with financial stock data on daily base.
I calculate in Excel INPUT Data and load them manually (File xlsreader) into a trained KNIME GBM(Regression) Model for prediction. I let run the workflow manually and give out the prediction values into an xlsx-Excel-file again. Now I can check the predictions for signal.
My problem now is, that all these efforts took to much time. Principially I should have the model in Excel and than I can caluclate online (with realtime data) my Input values in order to decide if I should by a stock or not.
As you can imagine I am not talking about 1 stock I have to check, we are talkin about a few hundred stocks.
So, if there would be a way, tho connect excel realtime directly with kime with regard to input and output (production), this would be of interest for me.

Many thanks

Bernd

@Bernd

For me this sounds like a job just perfect for KNIME. Process large quantities of data and apply models.

If you use loops or parallel processing you could do a lot of operations in a short time. You could take a look at this free book explaining functions between excel an KNIME.

https://www.knime.com/knimepress/download-from-excel-to-knime

KNIME server would allow for automation but you might also try cron jobs or windows task manager to schedule tasks.

And also you would not be limited to PMML compatible models. Thinking deep learning.

1 Like