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

undefined subtrees. To generate such a tree, the construction and pruning oper- ations are integrated in order to find a “stable ...

6.2 CLASSIFICATION RULES 209 at each point the lowest-entropy sibling is chosen for expansion: node 3 between stages (a) and (b) ...

method that builds a full decision tree each time. The gain decreases as more pruning takes place. For datasets with numeric att ...

6.2 CLASSIFICATION RULES 211 It is used to split the training data into two subsets: one containing all instances for which the ...

examples that satisfy it divided by the number that satisfy its condition but not its conclusion. For example, the condition in ...

6.2 CLASSIFICATION RULES 213 Of the rules for these instances that do not predict the default class Iris setosa, the best is if ...

The idea of incremental reduced-error pruning is due to Fürnkranz and Widmer (1994) and forms the basis for fast and effective r ...

6.3 EXTENDING LINEAR MODELS 215 constructed in the new space can represent a nonlinear decision boundary in the original space. ...

classes are linearly separable; that is, there is a hyperplane in instance space that classifies all training instances correctl ...

6.3 EXTENDING LINEAR MODELS 217 in the two-attribute case, where a 1 and a 2 are the attribute values, and there are three weigh ...

maximum margin hyperplane is relatively stable: it only moves if training instances are added or deleted that are support vector ...

6.3 EXTENDING LINEAR MODELS 219 way of choosing the value ofnis to start with 1 (a linear model) and incre- ment it until the es ...

tions’ absolute error instead of the squared error used in linear regression. (There are, however, versions of the algorithm tha ...

6.3 EXTENDING LINEAR MODELS 221 0 2 4 6 8 10 0 2 4 6 8 10 class (a) attribute 0 2 4 6 8 10 0 2 4 6 8 10 class (b) attribute 0 2 ...

straint, and all training instances become support vectors. Conversely, ifeis large enough that the tube can enclose all the dat ...

6.3 EXTENDING LINEAR MODELS 223 This rings a bell! A similar expression for support vector machines enabled the use of kernels. ...

learning. Previously, neural network proponents used a different approach for nonlinear classification: they connected many simp ...

C AB 1 ("bias") 1 attribute ("bias") a 1 attribute a 2 –1.5 1 1 1 1 –1 –1 –0.5 1.5 (e) 1 1 – 1.5 attribute a 1 attribute a 2 1 ( ...

tron, the input layer has an additional constant input called the bias.However, the third unit does not have any connections to ...

6.3 EXTENDING LINEAR MODELS 227 Backpropagation Suppose that we have some data and seek a multilayer perceptron that is an accur ...