Hitachi Vantara Pentaho Community Forums
Results 1 to 4 of 4

Thread: Mismatch in output prediction and Correctly Classified Instances

  1. #1
    Join Date
    Jan 2016
    Posts
    2

    Default Mismatch in output prediction and Correctly Classified Instances

    Hi All,

    I am using BatchPredictorVote that has 2 classifiers with majority voting for which I am getting results as mentioned below:
    Correctly Classified Instances 17 89.4737 % and
    Incorrectly Classified Instances 2 10.5263 %.

    However when I check my output predictions of test data, I see only 14 predictions correct and 5 predictions wrong. I have attached the screenshot of the same. Am I missing something?

    Thanks,
    Sanjay
    Attached Images Attached Images  

  2. #2
    Join Date
    Aug 2006
    Posts
    1,741

    Default

    Hi,

    This classifier is intended only for internal use in the distributed Weka environment. As such, it doesn't yet implement batch prediction for all the combination strategies (only average of probabilities is implemented). There is no reason to use this classifier over standard Vote. I'll make sure that it is not visible in the GUIs in a future release.

    Cheers,
    Mark.

  3. #3
    Join Date
    Jan 2016
    Posts
    2

    Default

    Quote Originally Posted by Mark View Post
    Hi,

    This classifier is intended only for internal use in the distributed Weka environment. As such, it doesn't yet implement batch prediction for all the combination strategies (only average of probabilities is implemented). There is no reason to use this classifier over standard Vote. I'll make sure that it is not visible in the GUIs in a future release.

    Cheers,
    Mark.
    Hi Mark,

    Thanks for the response. With standard Vote classifier also it is same behavior. It mentions correctly classified instances as 17 but in output prediction only 13 are correct. See attached screenshot.

    Thanks,
    Sanjay
    Attached Images Attached Images  

  4. #4
    Join Date
    Aug 2006
    Posts
    1,741

    Default

    Bummer :-) Under the majority voting combination scheme, Vote breaks ties in the counts for the votes randomly. As the same Vote object is used to produce the evaluation predictions, and then the predictions again for outputting the individual predictions, the predictions for cases where there are ties can differ. I've just committed a fix to our subversion repository where Vote will now use the probability averaging combination strategy to break ties when using majority vote (i.e. it will look at the maximum probability in the distribution in this case).

    You can get the fix in the next nightly snapshots of Weka.

    Thanks for bringing this to our attention.

    Cheers,
    Mark.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
Privacy Policy | Legal Notices | Safe Harbor Privacy Policy

Copyright © 2005 - 2019 Hitachi Vantara Corporation. All Rights Reserved.