Compare ROCs when a transform has to be applied on the data of one of the models

moudar981moudar981 Member Posts: 5 Contributor II
edited November 2018 in Help
We can use CompareROC to plot two or more ROCs, I need to compare the performance of svm with svm after applying LDA on the data. As LDA is not in rapid miner, I compute it externally and read the result into rapidminer using ARFF Reader. Is there a way to plot ROC comparison in this case?

Answers

  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi,

    LDA is indeed part of RapidMiner, but not in the abbreviated form: Instead it is called Linear Discriminant Analysis.
    If you still want to use your external tool, this can be compared, if it generates confidence values for the class predictions. These are used to generate the ROC plot. If it does all you need to do is to set the role of the respective attributes to "confidence_<class>".

    Greetings,
      Sebastian
  • moudar981moudar981 Member Posts: 5 Contributor II
    thank Sebastian,
    Yes, I know LDA is part of RapidMiner, but it is a classifier, while I need it as features transform, and then use another classfier on the transformed features.

    Firas
  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi,
    you can use the confidence attributes as the transformed features. :)

    Greetings,
      Sebastian
Sign In or Register to comment.