Hitachi Vantara Pentaho Community Forums
Results 1 to 3 of 3

Thread: Doubts

  1. #1
    Join Date
    Oct 2015
    Posts
    11

    Default Doubts

    Hello friends, I have no doubt in two respects in relation to the PCA (Principal Component Analysis) and I would like to help me.The first: I used PCA filter in my database, but the classification rate was worse than when done with the original base. I used various algorithms to make sure and even then the rates were lower. I wonder why? and will always be so?The second: I have problem with the review process. To be more accurate, use the PCA filter process in Weka on a base of 4500 attributes, the program does, however when I do the same procedure with a 7000 attributes in the database, the program after an error, not performs. I've upped maxheap to -Xmx4000m, but without success. COuld you help me?



    Thank you! Sorry for English

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

    Default

    The default in PCA is to retain enough principal components to account for 95% of the variance in the original data. Try increasing that to 100%. PCA has runtime that is cubic in the number of attributes, so is not a good choice when there are large numbers of attributes. If you are looking to reduce the number of attributes via attribute selection then try a simple attribute ranking scheme, such as info gain ranking - this has runtime that is linear in the number of attributes. For something a little more powerful you could try Weka's RankSearch method combined with the ClassifierSubsetEval evaluator. This is a wrapper scheme that also has runtime that is linear in the number of attributes (not taking the computational complexity of the base learner into account).

    Cheers,
    Mark.

  3. #3
    Join Date
    Oct 2015
    Posts
    11

    Default Doubts

    Thank you Mark!
    Last edited by Jovani; 04-12-2016 at 02:31 PM.

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.