unable to load attributes using AttributeConstructionLoader

Legacy User
Legacy User New Altair Community Member
edited November 2024 in Community Q&A
Hi,

I have a problem with loading my attributes using the AttributeConstructionLoader. Firstly, I generated my attributes in another rapidminer file using genetic algorithm and i used the AttributeConstructionWriter to write all the attributes in an *.att file. I checked the file, it has 19 attributes.

Then I opened a new rapidminer file. I used CSVExampleSource to load my dataset (the same one I used for generating attributes; I used the same dataset because I was just trying out...), then I placed the AttributeConstructionLoader under root and selected my attribute file to load. When I run the process, the outcome was that there were no attributes left.

Does anyone have any idea what might have went wrong?

Thanks
Tagged:

Answers

  • IngoRM
    IngoRM New Altair Community Member
    Hello,

    the operator AttributeConstructionLoader has a parameter called "keep_all". Did you try to activate it?

    The sample processes "05_Features/19_YAGGA.xml" and "05_Features/20_YAGGAResultAttributeSetting.xml" should demonstrate how the AttributeConstructionLoader can be used.

    Cheers,
    Ingo
  • Legacy User
    Legacy User New Altair Community Member
    Hi,

    Thanks for your advice. From the examples I have found AttributesWeightWriter and AttributeWeightsLoader to be useful for my purpose.

    I am not sure how the AttributesConstructionLoader works, it seems that if I simply used the AttributesConstructionWriter to write the attributes, and when I load them, there would be no attributes left, and if keep all == true, all the attributes would remain, so in the end I cannot only retain the attributes I want.

    Thanks
  • IngoRM
    IngoRM New Altair Community Member
    Yes, that is true. The AttributeConstruction... operators are only for constructions, not for feature selection. For this purpose, the AttributeWeights... operators have to be used.

    Cheers,
    Ingo