none
Is there a way to get the permutation of Features for each testing data? RRS feed

  • Question

  • Permutation Feature Importance module provides permutation of whole testing data at once, but i am trying to get the features who contributed most in classifying each testing data. Is there any way to do that?

    The dataset is having 5 value columns and 1 target column

    The purpose of doing this is, i have to show to the user that these fields gave major contribution and we can gain their trust that these fields you have entered are responsible for the result. 

    Thanks in advance. 

    Friday, September 20, 2019 5:06 AM

All replies

  • Hi,


    Can you please clarify the difference between whole testing data and each testing data from your post? Based on my understanding of your question, you would need to create separate experiments for each testing dataset. Once you connect the Permutation Feature Importance module to the Train Model and run the experiment, you can simply right click on Permutation Feature Importance module, select Feature Importance > Visualize. You should see a score for each feature (ranked in descending order). Important features would result in higher importance scores and vice versa. Please let me know if this helps, or kindly explain your question in more detail so I can further assist you. Thanks.

     

    Regards,

    Azure CXP Community.

    If a post helps to resolve your issue, please click "Mark as Answer" and/or "Vote as helpful". By marking a post as Answered and/or Helpful, you help others find the answer faster.  Thanks.


    Friday, September 20, 2019 10:04 PM
    Moderator
  • Thank you for your response,

    In my experiment, whole testing data  means a data set of 1000 or more entries (for ex: 'abc' is a variable who contains 1000 records in it) where as each testing data means a single record inside the dataset (abc[0] is a single record).

    I have used Permutation Feature Importance module but it gives result for complete dataset i.e. for 'abc' but i need this  Feature importance score for each record i.e for every abc[i] where 0>i>999.

    I hope this time you get my problem. Still  if have any other confusion then ask me .


    Sunday, September 22, 2019 12:39 PM
  • Hi,

     

    Thank you for clarifying your question. However, I don’t think Permutation Feature Importance works the way you're intending for it to work. It is not designed to compute features importance scores given a dataset record, but rather it computes the permutation feature importance scores of feature variables given a trained model and a test dataset. Please feel free to refer to this document for more information, and let me know if you have further questions. Thanks.

     

     

    Regards,

    Azure CXP Community.

    If a post helps to resolve your issue, please click "Mark as Answer" and/or "Vote as helpful". By marking a post as Answered and/or Helpful, you help others find the answer faster.  Thanks.

    Tuesday, September 24, 2019 12:22 AM
    Moderator
  • Hi,

    Thanks for responding. My issue isn't resolved yet, so thinking of creating  my own customized model

    Tuesday, September 24, 2019 4:26 AM
  • Hi,

     

    Okay, let me know if I can be of any more help. In the mean time, please feel free to submit product feedback through the Azure ML User Voice forum where you and others can vote for that functionality. Thanks.

     

     

    Regards,

    Azure CXP Community.

    If a post helps to resolve your issue, please click "Mark as Answer" and/or "Vote as helpful". By marking a post as Answered and/or Helpful, you help others find the answer faster.  Thanks.

    Tuesday, September 24, 2019 5:10 PM
    Moderator