How I can de-normalize a predictive value generated using the NN?

Hello all-

I trained a NN using as input a set of normalized data, the NN was trained…but the predictive values were delivered in a normalize form. How I can de-normalize these predictive values?

Best Regards,


Dear Marmota,

when you where normalizing the data the Normalizer node has two outports. The normalized data (port) and the model. This model port contains all information you need for denormalizing. Feed it and the data into the denormalization node. Please note that the column names need to be the same as they were before the normalization.

Best, Iris

1 Like

I can't seem to figure this out either. 

Iris the normalizer model doesn't include any information for the predicted values. In my case im normalizing 'CA binding' then creating predictions for 'CA binding' my new column 'CA binding predicted' isn't getting denormalized. Is there a way to apply the 'CA binding' normalization models to my 'CA binding Predicted'?


Just rename the column CA binding then denormalize it. You may have to use to seperate denormalization nodes if you need both values. You  can just rename it again after denormalization. Unfortunately this is the easiest way to do this.


1 Like


I have the same concern. Denormalizer only inverse normalize the input label, not predicted label due to different column name

Dear all,

Please find a minimalist workflow which explains how to deal with normalization and denormalization of data when using a NN, although it can be adapted to any regression problem.

Hope this helps.

Have a nice worker’s day !

20210501 Pikairos How I can de-normalize a predictive value generated using the NN.knwf (517.1 KB)



Hello all,

see here if not working with NN (meaning you can’t run this workflow) but still have a need to denormalize prediction column: