Binomial Classification Question

GhostriderGhostrider Member Posts: 60 Contributor II
Say I'm trying to do binomial classification.  I know that if I try to classify every single example, my accuracy won't be too good.  But what matters is that the examples which have high classification confidence are correctly classified.  So I don't need to know the classification for all examples, I just need to know which examples the learning algorithm can classify with a high level of confidence.  Is there a learning algorithm in RapidMiner that supports this?

Answers

  • IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder
    Hi,

    yes, actually all classification schemes in RapidMiner deliver also the confidences when their model is applied to new data. There are even some post-processing operators which can be used for rescaling those confidences so that they better match actual probabilities and there also operators for filtering out the uncertain cases (which is often very useful in practice).

    Cheers,
    Ingo
Sign In or Register to comment.