Data Mining Techniques for Segmentation - Data Mining Techniques in CRM: Inside Customer Segmentation - page 128

Figure 3.29 IBM SPSS Modeler recommended CHAID Model options.
Things to bear in mind about PCA:
• Numerous different criteria can be taken into account for determining the
number of components to be extracted. Most of them relate to the amount of
original information jointly and individually accounted by the components. The
most important criterion, though, must be the interpretability of the components
and whether they make sense from a business point of view.
• Component interpretation is critical since the derived components will be
used in subsequent tasks. This usually includes an examination of the loadings
(correlations) of the original inputs on components, preferably after applying a
Varimax rotation.
• Derived component scores can be used to represent and to substitute for the
original fields. Provided of course that they are derived by a properly developed
model, they can be used to simplify and even refine subsequent modeling
tasks, including clustering, since they equally represent all the underlying data
dimensions.