(Section 6.1, page 190). It deals with missing values by splitting instances into
pieces, as C4.5 does. You can set the minimum number of instances per leaf,
maximum tree depth (useful when boosting trees), minimum proportion of
training set variance for a split (numeric classes only), and number of folds for
pruning.
NBTreeis a hybrid between decision trees and Naïve Bayes. It creates trees
whose leaves are Naïve Bayes classifiers for the instances that reach the leaf. When
constructing the tree, cross-validation is used to decide whether a node should
be split further or a Naïve Bayes model should be used instead (Kohavi 1996).
M5Pis the model tree learner described in Section 6.5.LMTbuilds logistic
model trees (Section 7.5, page 331).LMTcan deal with binary and multiclass
target variables, numeric and nominal attributes, and missing values. When
fitting the logistic regression functions at a node, it uses cross-validation to
determine how many iterations to run just once and employs the same number
throughout the tree instead of cross-validating at every node. This heuristic
(which you can switch off ) improves the run time considerably, with little effect
on accuracy. Alternatively, you can set the number of boosting iterations to be
used throughout the tree. Normally, it is the misclassification error that cross-
validation minimizes, but the root mean-squared error of the probabilities can
be chosen instead. The splitting criterion can be based on C4.5’s information
gain (the default) or on the LogitBoost residuals, striving to improve the purity
of the residuals.
ADTreebuilds an alternating decision tree using boosting (Section 7.5, pages
329–331) and is optimized for two-class problems. The number of boosting
iterations is a parameter that can be tuned to suit the dataset and the desired
complexity–accuracy tradeoff. Each iteration adds three nodes to the tree (one
split node and two prediction nodes) unless nodes can be merged. The default
search method is exhaustive search (Expand all paths); the others are heuristics and
are much faster. You can determine whether to save instance data for visualization.Rules
Table 10.5 shows many methods for generating rules.DecisionTablebuilds a
decision table majority classifier (Section 7.1, page 295). It evaluates feature
subsets using best-first search and can use cross-validation for evaluation
(Kohavi 1995b). An option uses the nearest-neighbor method to determine the
class for each instance that is not covered by a decision table entry, instead of
the table’s global majority, based on the same set of features.OneRis the 1R
classifier (Section 4.1) with one parameter: the minimum bucket size for dis-
cretization.ConjunctiveRulelearns a single rule that predicts either a numeric
or a nominal class value. Uncovered test instances are assigned the default class408 CHAPTER 10 | THE EXPLORER