Test Set Inclusion Method

When a test set is to be incorporated into the training process, Qnet must select which patterns to set aside for testing. You may instruct Qnet to select the patterns randomly from the training file or simply use the beginning or ending specified number of patterns from the file. If the training data is in some type of systematic order, selecting the “random” method will normally offer the best results. If the training cases are not organized in some particular way that would effect the training/testing results, selecting the points from the beginning or ending of the file is also appropriate. For the “random” method, the points will be selected once. Subsequent training session using the same input data with the same number of test cases will result in the same random cases being used. If the total number of training cases changes or if you later select a different number of test set cases, a new random selection will occur.