"Remove an attribute for model training and include them for visualization"

radone

I have ExampleSet which also contains attribute "DATE".
This attribute I do not want to use for model training but I would like to use it for result visualization .

Is there any way how achieve this? Or any other ideas?

My idea was to create two chains (OperatorChain (2) and OperatorChain, see below) and apply FeatureNameFilter (2) on the first chain, which removed DATE attribute. Unfortunately, this attribute is removed also in the second testing chain.

Root[1] (Process)
  +- ExampleSource[1] (ExampleSource)
  +- SlidingWindowValidation[1] (SlidingWindowValidation)
      +- OperatorChain (2)[1] (OperatorChain)
      |  +- FeatureNameFilter (2)[1] (FeatureNameFilter)
      |  +- W-REPTree[0] (W-REPTree)
      +- OperatorChain[0] (OperatorChain)
        +- Applier[0] (ModelApplier)
        +- BinominalClassificationPerformance[0] (BinominalClassificationPerformance)



Thanks in advance for any ideas

radone

fischer

Hi,

you can use a ChangeAttributeRole to make the attribute special. E.g. set its role to "ignored". You can choose any name for this value. The attribute will stay in the example set, but the learner will ignore it (unless you set the role to "label" :-)).

Cheers,
Simon