There is a filter in Weka called "Add Classification". One can choose a classification algorithm and then can get the classification labels provided by the model generated by the algorithm, the distribution of the classes, and errors it is making. Then, one can add these information (classification labels, class distributions, and errors) into the existing feature/attribute set.
For instance, a dataset has 30 attributes (including class attribute) and 2 classes. Using this filter one can generate classification labels, 2 distributions, and classification errors. Adding them to the original dataset gives 34 attributes.
My question is: Does this method of using one classifier's labels, distribution and errors as additional features for another classifier have any particular name? I would like to know more about the efficiency of this method. Can anyone help me please?
