If the size of the testing data set is too small, the result might bias
toward one method over the other(s). How can I determine an appropriate
testing data size? Or, how can I judge if the size is too small, so that
the performance result should not be interpreted as one method is superior
to another?
Thanks
|