Use learned model to predict new test data
Previous  Top  Next

This chapter explains how to use the learned model to predict the output values for some test data tuples given only the input values.
All controls referred to are located on the tab sheet Use Model.

Load the test data file
Hit the Load button and select the test data file. This file must match the
test data file format. The number of data tuples and variables is displayed, if the file is loaded successfully.

Modify the prediction parameters
You may modify the
parameters used for testing by checking the Modify check boxes in the corresponding group box.

Note: If you uncheck the check box again, the values are restored from the ones, that were originally set to learn the model.

Set the simulation output file
Hit the button Set Output to set the simulation output file to which the predicted output values are written. This file is required, if the test data is without output.

Start the simulation
Hit the button Test.

Note: This button is only enabled if a test data file is loaded and, if this test data is without output, if a simulation output file is specified.

Test data prediction accuracy
If the test data file is with output, then the test data
prediction accuracy is displayed after the simulation is finished. Shown is the overall performance as well as the inside and the outside performance. The former is the performance of all predictions, that were made based on the rule-base component of the prediction mechanism, whereas the latter is the performance of all predictions, that were made based on the k-nearest-neighbor component.