Page 289 -

272 Chapter 8

techniques, and database technology, data mining detects hidden patterns and subtle
relationships in data and infers rules that allow the prediction of future results. Raw
data are analyzed to put forth a model that attempts to explain the observed patterns.
This model can then be used to predict future occurrences, and to forecast expected
outcomes (see fi gure 8.2 ).
A large number of inputs are required, usually over a signifi cant period of time, and
the types of models produced range from easy to almost impossible to understand.
Easy to understand models are decision trees, for example. Regression analyses are
moderately easy to understand and neural networks remain black boxes. The major
drawback of the black box models is that it becomes very diffi cult to hypothesize about
causal relationships (see fi gure 8.3 ).

Historical
data

If Then xxxx

Data
mining

If Then yyyy

Figure 8.2
Predictive models

Age

How well will the
Education
student perform on
the entrance exam?
Eye color
Model
Figure 8.3
Black box models