Platinum

Tutorial 6: Step 9 Classify Test Data

We now further test our classifier by predicting the classes of some samples which it has not already seen. These are in the Khan_test_data dataset. We have already filtered it, so we have a subset containing exactly those same genes we have just used to train the classifier.

Classify New Samples

1. Click the Filtered: keep {Tutorial 6 list} item under Khan_test_data in the Experiments navigator.

2. Click the Classify toolbar icon , or select Classify from the Predict menu, or right-click the item and select Classify from the shortcut menu. The Classify parameters dialog is displayed.

3. Set dialog parameters.

Parameter	Setting
Name	Type in a name for the variable which will contain the predicted classes of the test data. Predictions is a suitable name in this instance. (If you were planning on doing multiple different predictions, you might want to give it a more distinctive name.)
Description	If you wish, click in the field and type in a long, informative description to the prediction being carried out.
Classifier	This displays a subset of the Experiments navigator containing those classifiers that can be applied to the dataset. Click on the ANN: training classes item beneath the Khan_training_data heading (the classifier just trained).

4. Click OK. The Classify function is performed, a new variable is added to the dataset family, and a new Classify item (named Predictions) is added the Experiments navigator under the Filter Genes item.

If you have automatic visualizations enabled in your user preferences, the Classification plot is displayed showing the classification results.