Combining performance mesasures

ozgeozyazarozgeozyazar Member Posts: 21  Maven
edited June 14 in Help
hi !

ı am using cross validation and try to measure the performance of model. I need both classification and binomial performance results but I do not know now to combine them. 

Could you please help me ? 

Best Answer

  • varunm1varunm1 Posts: 563   Unicorn
    edited May 17 Solution Accepted
    Hello @ozgeozyazar

    Simple way is to use a multiply operator inside testing part of CV. Connect the apply model output to multiply and you can use the outputs of multiply operators to connect to the two performance operators.


Answers

  • ozgeozyazarozgeozyazar Member Posts: 21  Maven
    @varunm1

    Yes Sir, 

    It definately works.

    Thanks so much :smiley:
    varunm1
  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,172   Unicorn
    Another way is to connect the performance output of one operator to the input of another performance operator. This creates a combined performance output. All the performance operators allow you to do this.  Then you don't even need the Multiply operator.
    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
    varunm1ozgeozyazar
  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 2,059  RM Data Scientist
    Hi,
    Just because i ran into this: Are there more people around who want to change the main criterion of a performance vector after building it like @Telcontar120 described it? I ran into this like 3 times and i am close to add a new operator to Toolbox.

    There is a 3 line groovy solution though.

    BR,
    Martin
    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
    varunm1ozgeozyazar
  • varunm1varunm1 Member Posts: 563   Unicorn
    Till I saw @Telcontar120 solution, I am not aware that we can do this way. This is useful.

  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,172   Unicorn
    @mschmitz an operator sounds useful but also not a problem that personally I have faced very often in this setup---but sometimes in optimization scenarios it is also useful to be able to change the main performance criterion.
    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
    ozgeozyazar
Sign In or Register to comment.