Orange Forum • View topic - Testing of Training and Validation Data Set

Testing of Training and Validation Data Set

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

Testing of Training and Validation Data Set

Postby Thom » Wed May 31, 2006 3:10

I was hoping I might get some direction in setting up a schema that can take a seperate training and validation data sets and use the test learner widget to obtain various evaluation results for the test set and for the validation set to compare the results. Maybe a graphical representation of using the Logistical Regression Widget.

Also is there a way to get a list of the Log OR coefficents from the Nomogram instead of the graphical results?

Thanks For the Help,
Thom

Postby Blaz » Fri Sep 08, 2006 18:59

This is way overdue, but i guess you probably solved it yourself already. anyway (since the first part is simple): either you read the two data sets using the File widget, or you split one with Data Sampler. Then use Test Learners, like in the following schema:

Image

Channels btw Data Sampler and Test Learners are set as follows:

Image

In Test Learners you have to switch btw Test on Train Data and Test on Test Data to see the difference. This is not the most elegant, but with Test Learners being rather elemental, this is all you can do for now. Test Learners does not display the differences btw test/train data accuracies.

I did not understand how you would like to compare the performance of the classifiers on two or more samples using the Nomogram (they were not designed to do this).


Return to Questions & Support



cron