version of the pruned tree from C5.0 is shown below. The first
ode employed the partitioning rule Petal.Length1.9 to partition
e space. This rule generated two subspaces. One of which was
he Setosa flowers. The other subspace was composed of the rest
es of flowers. The second branch node employed the partitioning
al.Width 1.7. This rule generated two subspaces. One was
d of 54 Versicolor of flowers and five Virginica flowers. The
s composed of 46 Virginica flowers and one Versicolor flower.
ength <= 1.9: setosa (50)
ength > 1.9:
al.Width <= 1.7: versicolor (54/5)
al.Width > 1.7: virginica (46/1)
Fig. 3.42. A pruned C50 tree constructed for the Iris data.
provides a bootstrap approach for delivering a robust decision tree.
ten trials of C5.0 were generated using the following code. There
en trees. Figure 3.43 shows two of ten trials. It can be seen that
w some difference.
C5.0(Species ~.,data=iris,trials=10)
