4.1. The MC4 Learner
Decision tree learners like MC4 are good at dealing
with relatively static situations, that is, situations in
which the relationships between the various attributes
are Wxed. We were not sure how true this was of the Spurs
team, and its performances, over the period being
examined. The overall classiWcation error of the MC4
learner for disjoint training and test data sets in the general
model was 69.81% and 61.35% for the expert chosen
data.