Hi Martin,

Martin Guetlein wrote:
> Hello All,
>
> just to make sure:
>
> We agreed on the parameter prediction_feature for creating models.
> This specifies which feature to predict in the test/training dataset.
>
> 'Prediction_feature' is not necessarily the feature of the predicted
> values in the prediction dataset (That is the case in Christoph's
> service, though).
>
>   
'prediction_feature' defines which variable is the dependent one, this
can be found in model ot:dependentVariables.

On the other hand, ot:predictedVariables defines where results are
stored.  IMHO IST services should change the current behaviour.

> The validation service has to access the field predictedVariables of
> the created model (as predictedVariable != prediction_feature, at
> least in Pantelis' service).
>   
predictedVariable should not be the same as 'dependentVariable' , unless
we want overwriting the 'observed' data.   This is specified in
opentox.owl and is implemented in TUM, NTUA and IDEA services.

Best regards,
Nina
> Right?
>
> Regards,
> Martin
>
>