Options

making ROC curve for cost sensitive learning

IatiiIatii Member Posts: 8 Contributor I
edited November 2018 in Help
Tagged:

Best Answer

  • Options
    yyhuangyyhuang Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 364 RM Data Scientist
    Solution Accepted

    Hi latii,

     

    I did not notice you were trying to "compare roc". To use the metacost togehter with compare ROCs, you simply put your learners inside metacost nest, and put metacost inside 'compare ROCs'.

     

    compareROC.PNGInsideCompareROC.PNGInsideMetaCost.PNG

    The ROCs plot:

    ROCs.PNG

     

    The compareROC.rmp is zipped and attached. Feel free to use it.

Answers

  • Options
    yyhuangyyhuang Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 364 RM Data Scientist

    Hi,

     

    The ROC curve can be generated using 'performance(Bionominal Classification)' operator.

     

    If you open the tutorial process for 'MetaCost' operator from the help page. Replace the 'performance(classification)'  with performance (Binominal Classfication), you can select accuracy, AUC, f measure, etc. for your performance outputs.

     

     

    metacost.PNGAUC.PNG

    In the results view for performanceVector, click AUC for the plot of ROC

    ROC.PNG

     

     The roc.rmp is attached for re-generating the plot. Please download and unzip it and import the process in your RM studio.

     

     

     

  • Options
    IatiiIatii Member Posts: 8 Contributor I
  • Options
    yyhuangyyhuang Administrator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 364 RM Data Scientist
    Yes, for your first question.

    What GA operator are you trying to use?
Sign In or Register to comment.