settings. Note that BuildSettings has a “ has a ” relationship with
AlgorithmSettings . As mining functions are broadly divided into
supervised and unsupervised functions, the SupervisedSettings base
interface is used to represent all the supervised functions. For exam-
ple, classification is a supervised type function, so ClassificationSet-
tings inherits from SupervisedSettings . There are no common settings
for unsupervised functions, because these functions are significantly
different from each other. As a result, the settings for unsupervised
mining functions directly inherit from BuildSettings . For example,
clustering is an unsupervised type function, so ClusteringSettings
inherit from BuildSettings .
Figure 8-5 depicts the objects related to algorithm settings. Since
algorithms are used in the context of a function, AlgorithmSettings are
always used as part of a BuildSettings object. AlgorithmSettings is the
base interface for all algorithm settings. SupervisedAlgorithmSettings is
the base interface for algorithms supporting supervised functions.
For example, the decision tree algorithm is used for supervised
functions so TreeSettings inherits from SupervisedAlgorithmSettings .
Although many algorithms that can support classification and regres-
sion accept the same settings, some algorithms, such as support vec-
tor machine (SVM), differ enough to warrant their own interfaces.
Algorithm settings objects and their relationships.