A Walk-through-guide for Using Decision Trees Software - Data Mining with Decision Trees: Theory and Applications - page 155

Fig. 10.3 Weka classifier panel.
(3) unpruned — A binary parameter that indicates whether the second
step of pruning is disabled.
(4) useLaplace — A binary parameter that indicates whether the posterior
probabilities in the leaves are smoothed based on Laplace correction.
In order to train the classifier, click start button (located in the right
side of the classifier window). Weka will run the training algorithm 11 times.
First it trains a classification tree using the entire dataset. This tree is
presented to the user. Then it performs the 10 folds cross-validation proce-
dure to evaluate the predictive performance of the classification tree. Note
that in the 10 folds cross-validation procedure the training set is randomly
partitioned into 10 disjoint instance subsets. Each subset is utilized once in
a test set and nine times in a training set. Figure 10.5 presents the resulted
window. The output sub-window presents various predictive performance
measures such as: accuracy, AUC and confusion matrix.
Scrolling up through the output sub-windows shows that the tree
classifier is presented as a text (see Figure 10.6). Indentation is used
to convey the tree topology. The information for each internal node