Orange Forum • View topic - HINT and cross-validation

HINT and cross-validation

A place to ask questions about methods in Orange and how they are used and other general support.

HINT and cross-validation

Postby Guest » Thu Sep 22, 2005 14:48

Hi,

I am trying to use Orange to repeat the experiments in Zupan et al. (1999) "Learning by discovering concept hierarchies" and to perform more experiments with my own data. I know the class HINT can be used to generate a hierarchy as a classifier (orngCI.HINT). But I haven't found a way to use this classifier with 10-ford cross validation method implemented in orngTest.crossValidation:

Code: Select all
>>> data = orange.ExampleTable("monks-1.tab")
>>> rootH = orngCI.HINT(type="complexity", boundSize=3)
>>> results = orngTest.crossValidation([rootH], data, folds=10)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "C:\Python23\lib\site-packages\orange\orngTest.py", line 160, in crossValidation
    return apply(testWithIndices, (learners, (examples, weight), indices, indicesrandseed, pps), argkw)
  File "C:\Python23\lib\site-packages\orange\orngTest.py", line 366, in testWithIndices
    cr = classifiers[cl](ex, orange.GetBoth)
orange.KernelException: 'orange.Domain': attribute 'e'' not found
>>>


Seems it doesn't recognize the new attributes constructed by HINT. One solution could be to do all the procedure manually and step by step, that is instead of using HINT, I use orngCI.FeatureByMinComplexity and then orngCI.addAttribute and so on. But I would like to know wheather there is any better sulution.

Thanks in advance

Return to Questions & Support



cron