"Attribute weights with Kennard Stone?"

keithkeith Member Posts: 157 Maven
edited May 2019 in Help
Does the KennardStoneSampling algorithm use attribute weights to scale the distances it computes?  If not, is it possible to add that option?

Thanks,
Keith
Tagged:

Answers

  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi Keith,
    unfortunately KennardStone uses only the unweighted euclidean distance. We will add it on our ToDo, but since there is a hell of work waiting, we can't promise that it is already in the next release...

    Greetings,
      Sebastian
  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi again,
    I nearly forgot to mention, that you could use the AttributeWeightsApplier to transform the numeric values of your attributes according to the attribute weights. This will enable you to use KennardStone with weighted distance calculation.


    Greetings,
      Sebastian
  • keithkeith Member Posts: 157 Maven
    Thanks for that suggestion.  It is possible to use AttributeWeightsApplier inside EvolutionaryWeighting?  I tried setting up a process inside EvolutionaryWeighting that applies the weights to the data, then runs Kennard Stone.  However, I'm getting an error saying that AttributeWeightsApplier doesn't have any attribute weights as an input.  Is there some trick to getting that to work?

    Thanks,
    Keith
  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi Keith,
    it's unnecessary to apply the weights within the evolutionaryWeighting, because the exampleset delievered to the inner operators is already weighted by the current weights.

    Greetings,
      Sebastian
Sign In or Register to comment.