Orange Forum • View topic - Problem with Test Learner -> Test on Test Data wont work

Problem with Test Learner -> Test on Test Data wont work

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

Problem with Test Learner -> Test on Test Data wont work

Postby Guest » Sun Jan 03, 2010 23:53

Hi everyone,

i try to train a classifier on one dataset and test it on separate one. Im using orange-win-w-python-snapshot-2010-01-03-py2.5.exe.

As train set i have chosen hayes-roth_learn.tab from orange and for testing hayes-roth_test.tab.

I tried following setups:
http://voyagememoirs.com/pharmine/2008/05/16/orange-part-vi-model-validation-using-cross-validation-andor-independent-validation-set/

and

http://www.ailab.si/orange/forum/viewtopic.php?t=827

Both of them give a F1 score of -1 for the test set and the classifier classifies everything in on class.

Is this a bug or am I doing something wrong?

Thanks in advance!

Postby Ales » Mon Jan 04, 2010 10:37

I cant reproduce this error. Do all of the classifiers act the same way? What are the specific classifier settings you are using?

Postby Guest » Mon Jan 04, 2010 13:57

I tested it with kNN and SVM and didn't change the settings. I downloaded orange 1.0 and there it works well.

Postby Guest » Tue Jan 05, 2010 23:39

Hi again,

the problem still exists. Even though the screenshot is from version 1, the result is the same in aforementioned version

Image

file loads adult.tab, data sampler selects 2000 examples, select attributes selects alle the attributes and the class is set correctly. This "path" ist connected to the data port of test learners.

the second path loads adult_sample.tab, selects the same attributes as above and is connected to the separate data port.

The result and the settigs are visible in the screenshot.
Setting test learners to test with crossvalidation works fine.
Is there any idea whats wrong here? What is my mistake?

Thanks in advance[/img]


Return to Questions & Support



cron