Information Technology Reference
In-Depth Information
Decision tree of tablet disintegration time
(reprinted and adapted from Shao et al., 2007a;
with permission from Elsevier)
Figure 5.9
decision tree. Data with numerical attributes are classifi ed
according to the similarity among them, and organized into
different classes. A local linear regression model is built for
each class and thus the global nonlinear relation is simulated
through a set of simple local linear regression models. The
second stage 'prunes' this tree by replacing the sub-trees
with linear regression functions wherever this is appropriate. In
order to validate the predictability of model trees, the correlation
coeffi cient was computed against the validation data (Figures 5.10
and 5.11).
Model trees are expressed as a collection of rules, where each
rule has an associated multivariate linear equation. Whenever a
situation matches a rule's conditions, the associated equation is
used to calculate the predicted value. Each rule also carries some
descriptive information that is useful for estimation of the rule's
Search WWH ::
Custom Search