Hitachi Vantara Pentaho Community Forums
Results 1 to 3 of 3

Thread: about the dataset of one-class libsvm in Weka

  1. #1
    Join Date
    Jul 2013
    Posts
    5

    Default about the dataset of one-class libsvm in Weka

    Hi,
    Currently I am doing some outlier detection, so I am playing with one-class libsvm. However, I just can not get it right.

    My training dataset comprises of some numeric attributes and an unary attribute(with the value "no" indicating "not outlier"). The testing set is similar.
    Then after I invoked one-class libsvm, it showed that the "correctly classified instance" and "incorrectly classified instance" are all 0!. By showing the outputs, I have seen all the outputs are all "?" (no prediction). Why? I correctly get the oneclassclassifer and SMO work to detect the outlier. I just can not create the correct dataset to get the one-class libsvm work

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

    Default

    Take a look at this weka list posting:

    http://list.waikato.ac.nz/pipermail/...er/038023.html

    Note that in newer versions of Weka the outlier instances predictions are represented by missing value ("?") instead of NaN.

    Cheers,
    Mark.

  3. #3
    Join Date
    Jul 2013
    Posts
    5

    Default

    Quote Originally Posted by Mark View Post
    Take a look at this weka list posting:

    http://list.waikato.ac.nz/pipermail/...er/038023.html

    Note that in newer versions of Weka the outlier instances predictions are represented by missing value ("?") instead of NaN.

    Cheers,
    Mark.
    Hi Mark,

    I have checked the post. That's exactly what I did. In my file, besides some numeric attributes, the last column is an unary attribute( @attribute outlier {no}). However, even using the training set itself as the testing set, after I invoke libsvm oneclass, all the predictions are "?"... PS: I use the latest 3.7 version

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.