Options

# Z-Trasformation

Dear listers,

I've build a model with some Z-transformed variables,

I should pre-transform the variables than apply the model, but the transformation is affected by the mean and the standard deviation of the new dataset values.

So If I apply the model and his rules on the new dataset, the prediction is not always the same, is related to the 'shape' of the new data.

Where I'm wrong?

Thanks for your attention.

Cristiano.

I've build a model with some Z-transformed variables,

I should pre-transform the variables than apply the model, but the transformation is affected by the mean and the standard deviation of the new dataset values.

So If I apply the model and his rules on the new dataset, the prediction is not always the same, is related to the 'shape' of the new data.

Where I'm wrong?

Thanks for your attention.

Cristiano.

0

## Answers

2,531Unicornthe Normalization operator will return a preprocessing model. This contains the 'shape' (mean and std. deviation) of the original data and can be applied on new data.

Simply apply the preprocessing model first, the the prediction model...

Greetings,

Sebastian

16Contributor IIThanks for your support!

So I should use 2 'Apply Model' , the first is for preprocessing and the second for the model,

In the first node:

the model input is the proprocessing output of the normalized original data

data: normalized new data

In the second node:

the model input is the original model

data: labeled data of the first node(normalized new data)

Is it correct?

Thanks again for your attention.

Cristiano.

2,531Unicorn16Contributor IIwhen I read the content of preprocessing model I see:

Normalize 33 attributes to mean 0 and variance 1.

.Using var1 --> mean 0.1

... 28 more attributes ..

How I see other normalize parameter's attributes?

Thanks for your support.

Cristiano.

16Contributor II2,531Unicornwhich other parameters?

Greetings,

Sebastian

16Contributor IIJust mean and variance.

C.

2,531Unicornthey aren't shown? Hm. Please make a feature request for that. Normally this should be displayable, since I saw that it is saved in the models.

Greetings,

Sebastian