0

Is it possible to create custom fold for Cross Validation in Weka. What I want is to have that each fold contains only the records that have a certain value for an attribute (e.g ZIP code).Is this possible?

Charles
  • 50,943
  • 13
  • 104
  • 142
Titus Pullo
  • 3,751
  • 15
  • 45
  • 65

1 Answers1

0

I think your best for achieving this is using something like R and the Weka interface RWeka. Weka doesn't allow you to create custom folds, so you have to do that another way and assemble the train/test splits yourself before passing them to Weka. This is relatively easily done in R.

Lars Kotthoff
  • 107,425
  • 16
  • 204
  • 204
  • But do I have to create my classifier using API and code? Because I need a quite complicate nested classifier (AttributeSelectd Classifier + FilteredClassifier with Oversampling/Undersampling filters +CostSensitiveClassifier + J48) so if I have to recode everything without a GUI looks quite hard! – Titus Pullo May 12 '12 at 17:00
  • You don't have to create the classifier yourself, the RWeka API allows you to call Weka directly. You will however have to process the data yourself, i.e. create folds and call filters etc before calling the final classifier. – Lars Kotthoff May 12 '12 at 17:08