3.2 Overview
At first, Section 3.3 will present the experimental setup, the datasets and classifiers considered and details on significance tests and alpha-confidence levels.
In Section 3.4 we address the choice of base classifiers, using MLR as meta
classifier as proposed by (Ting & Witten, 1999). We show that it is quite hard
to significantly beat the trivial choice of using all available base classifiers and