Orange Forum • View topic - Accuracy of the classifier

Accuracy of the classifier

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

Accuracy of the classifier

Postby veeraganesh » Mon Aug 13, 2007 15:28

after building the tree with training data when we pass the test data, is there any way we can check the accuracy of the classifier.For instance, the test data got 5 records classified as 4 to class A and 1 to class b by the classifier.Where as actually 3 belong to Class A and 2 belong to class B.So we can say that the accuracy of the classifier is 80%

Postby Blaz » Wed Aug 15, 2007 13:35

Please see Testing and Evaluating Your Classifiers in documentation. If you have a separate training and test set, all you should do is replace a call for orngTest.crossValidation with

Code: Select all
learnAndTestOnTestData(learners, trainset, testset)

where learners is your list of learners, trainset a set of training instances, and testset a data set for testing.

See also documentation for orngTest and orngStat.

In widgets, you should use Test Learners from Evaluation pane.

Return to Questions & Support