Result of Auto Model

sara20sara20 Member Posts: 110 Unicorn
edited July 2020 in Help
Hello all RM friends

Why in Auto Model in the last part some times we have only Accuracy? Why some times we don't have recall, precision, AUC and...?


Thank you in advance for your help
Sara
Tagged:

Answers

  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 1,195 Unicorn
    Hi @sara20,

    I think it is because in the first case, you have a multi-class classification problem and in the second case you have a binary classification problem (where you can calculate, in addition of the classification error and the accuracy,  the AUC, sensitivity etc .)

    Regards,

    Lionel 
  • sara20sara20 Member Posts: 110 Unicorn
    edited July 2020
    @lionelderkrikor

    Hello 

    Thank you for the answer but I have two datasets which they are the same. Both of them are numeric but it is really strange for the result and metrics( both of them are very normal datas). So what is the reason for that?

  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 1,195 Unicorn
    @sara20

    Can you specify the type of your label in both cases ?

    Regards,

    Lionel
  • sara20sara20 Member Posts: 110 Unicorn
    edited July 2020
    @lionelderkrikor

    Both of them have Nominal labels. َAlso yesterday I up date my RM to version 9.7. In version 9.6 I didn't have this problem. Is it the problem for version 9.7? If yes so how can I change it to version 9.6? or salve the problem?

    Thank you

  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 1,195 Unicorn
    @sara20

    In order we can reproduce what you observe, can you share : 

     - the two datasets
     - and specify in both cases the label

    Regards,

    Lionel

  • sara20sara20 Member Posts: 110 Unicorn
    edited July 2020
    @lionelderkrikor

    Thank you very much for following the problem but unfortunately the data is not mine and I can not share it. Can I change the process in Auto Model in order to fix the problem?imageimageimage
  • Telcontar120Telcontar120 Moderator, RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 1,635 Unicorn
    Even if both labels are nominal type, if in one dataset there are only two values of the label it will be treated as binominal and thus as @lionelderkrikor explained you will have additional metrics available like AUC.  But if there are more then two values of the label, it will be treated as polynominal and then you will have only accuracy.  You can always check this by taking the output of Auto Model and opening the underlying process and then checking the data type and the data values for your label.
    Brian T.
    Lindon Ventures 
    Data Science Consulting from Certified RapidMiner Experts
Sign In or Register to comment.