Data Mining: Practical Machine Learning Tools and Techniques, Second Edition

10.4 LEARNING ALGORITHMS 405

attributes using supervised discretization.NaiveBayesUpdateableis an incre-

mental version that processes one instance at a time; it can use a kernel esti-

mator but not discretization. NaiveBayesMultinomial implements the

multinomial Bayes classifier (Section 4.2, page 95).ComplementNaiveBayes

builds a Complement Naïve Bayes classifier as described by Rennie et al. (2003)

(the TF ¥IDF and length normalization transforms used in this paper can be

performed using the StringToWordVectorfilter).

AODE (averaged, one-dependence estimators)is a Bayesian method that aver-

ages over a space of alternative Bayesian models that have weaker independence

assumptions than Naïve Bayes (Webb et al., 2005). The algorithm may yield

more accurate classification than Naïve Bayes on datasets with nonindependent

attributes.

BayesNetlearns Bayesian networks under the assumptions made in Section

6.7: nominal attributes (numeric ones are prediscretized) and no missing values

(any such values are replaced globally). There are two different algorithms for

estimating the conditional probability tables of the network. Search is done

using K2 or the TAN algorithm (Section 6.7) or more sophisticated methods

based on hill-climbing, simulated annealing, tabu search, and genetic algo-

rithms. Optionally, search speed can be improved using AD trees (Section 6.7).

There is also an algorithm that uses conditional independence tests to learn the

structure of the network; alternatively, the network structure can be loaded from

an XML (extensible markup language) file. More details on the implementation

of Bayesian networks in Weka can be found in Bouckaert (2004).

You can observe the network structure by right-clicking the history item and

selecting Visualize graph.Figure 10.18(a) shows the graph for the nominal

version of the weather data, which in fact corresponds to the Naïve Bayes result

Table 10.5 (continued)

Name Function

SMOreg Sequential minimal optimization algorithm for support
vector regression
VotedPerceptron Voted perceptron algorithm
Winnow Mistake-driven perceptron with multiplicative updates
Lazy IB1 Basic nearest-neighbor instance-based learner
IBk k-nearest-neighbor classifier
KStar Nearest neighbor with generalized distance function
LBR Lazy Bayesian Rules classifier
LWL General algorithm for locally weighted learning
Misc. Hyperpipes Extremely simple, fast learner based on
hypervolumes in instance space
VFI Voting feature intervals method, simple and fast