Transpose Output correctly for comparison

Dear KNIME Community,

The “Keras Network Executer” Node gives me 24 predictions (one for each hours of the day) for every test sample.

It is however all in one row.

My original training target values are in a Column. The thing that ties them together is the testing sample number (which is in both).

For readability it is nice to have then side by side so that I can see how close my predictions are to the real values for the testing sample.

How do I move them side by side. I would also appreciate general comments on how to evaluate my test findings.

Much, much appreciated,
Leon

To transform your first table, you could use the Unpivoting node, retaining the Testing Sample column and setting all your predictions as value columns.

4 Likes

Thanks Scott,

That helps soooooo much!

Do you have any other general comments on comparing predicted results real targets?

Thanks a lot,
Leon

Assuming you have the rows sorted correctly after transformation, you could use the Numeric Scorer node to calculate summary statistics (RMSE, etc).

1 Like

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