user manual

34
Chapter 4
have been resolved adequately. Similarly, the evaluation phase can lead you to reevaluate your
original busi
ness u ndersta nding, and you may decide that you have been trying to answer the
wrong question. At this point, you can revise your business understanding and proceed through
the rest o f the process again with a better target in mind.
The second ke
y point is the itera tive nature of data mining. You will rarely, if ever, simply
plan a data mining project, c omplete it, and then pack up your data and go home. Data mining to
address your custome r s’ demands is an ongoing endeavor. The knowledge gained from one cycle
of data minin
g will a lmost invariabl y lead to new questions, new issues, and new opportunities
to identify and meet your custo mers’ needs. Those new questions, issues, and opport unities can
usually be addressed by mining your data onc e again. This process of min ing and identifyin g new
opportunit
ies should become part of the way you thi nk about your b usiness and a cornerston e o f
your overall business strategy.
This introduction provides only a brief overview of the CRISP- DM process mo del. For
complete de
tails on the model, consult the following resources:
 The CRISP-DM Guide, which can be accessed along with other documentation from the
\Documentation folde r on the installation disk.
 The CRIS P-DM Help system, available from the Start me nu or by clicking CRISP-DM Help on
the Help menu in IBM® SPSS® Mo deler.
Types of Models
IBM® SPSS ® Modeler offers a variety of modeling methods taken from machine learning,
artificial intelligence, and statistics. The methods available on the Modelin g palette allow you
to deriv e new information from your data and to develop predictive models. Each me thod has
certain strengths and is best suited for partic ular types of problems.
The SPSS Modeler Applications Guide provides ex amples f or many of these methods, along
with a general introduction to the modeling process. This guide is available as an online tutorial,
and also in P DF format. For more information, see the topic Application Examples in Chapter 1
on p. 5.
Modeli ng methods are divided into th r ee categori es:
 Classification
 Associa tion
 Segmentation
Classification Models
Classification models u se th e values of one or more input fields to pre dict th e value of one or
more output, or target, fields . So me examples of these techniques are: decisio n tre es (C&R Tree,
QUEST, CHAID and C5.0 algorithms), regression (linear, logistic, generali zed linear, and Cox
regress ion algorithms), neural networks, support vector machines, and Bayesian ne tworks.
Classification models hel ps organizations to predict a known result, such as whether a customer
will bu y or leave or wheth er a transaction fits a known pattern of fraud. Modeling te ch niques
include machine learning, rule induction, subgroup identification, statistical methods, a nd multiple
model generation.