🎉 🎉 RAPIDMINER 9.10 IS OUT!!! 🎉🎉

Download the latest version helping analytics teams accelerate time-to-value for streaming and IIOT use cases.



mbukombuko Member Posts: 10 Contributor I
edited July 2019 in Help

I am a student and try to understand the RapidMiner output after applying a trained model (e.g. multiclass with SVM, KNN, NB). I am getting the prediction and for each possible class/label the confidence value. But I don't know how to interpret these confidence values and how their calculation is defined.

Do you know, if there are definitions anywhere of the calculation and interpretation of the confidence values for each classification algorithm (e.g. SVM, KNN, NB)?

Best regards



  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 3,055  RM Data Scientist

    since the mentioned algorithms are standard algorithms you can search easily on the web. As a good quoteable, mathematical source i can recommend Hastie, Tibsheirani, Friedmann - The elements of statistical learning. But be aware - math dragons are around the corner.

    For k-NN: It is simply the (weighted) fraction of examples belonging to the class.
    NB: The result of the naive bayes formula. See: https://en.wikipedia.org/wiki/Naive_Bayes_classifier the one in RM is actually the gaussian one.
    For SVM: i recommend the book above.. Quite some fun with lagrange multipliers. In the end classification is only a scalar product. i think this is simply mapped to [0,1]

    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
  • mbukombuko Member Posts: 10 Contributor I
    Thank you for your recommendations, I will  examine it. In my case there are also negative values for the confidence values of the SVM classification, that is why I was a bit surprised. The relationship and description of the confidence values of these algorithms I  have not seen in any literature or website yet.

    Best regards
Sign In or Register to comment.