A set of test cases where the outcome is already known can be used either to
train or to measure the quality the classifier. It is a bad idea to use exactly the same
set of test cases to train as well as to measure the quality of the classifier, since
the classifier has already seen the correct classification of the test cases during
training; this can lead to artificially high measures of quality. The quality of a
classifier must therefore be measured on test cases that have not been seen during
training.