Practical Problem Solving - Java Data Mining: Strategy, Standard, and Practice - page 327

PurchaseA [Distance: 0.1954]

50000

40000

30000

20000

3

10000

1
2
3

2

0

1

Figure 12-5 Graphical profiling of one cluster with distance

information.

deviations of a continuous attribute in both cases (in or out of the
specified cluster).

12.3.4 Scenario 3 Conclusion

We have shown in this section that, in conjunction with JDBC, the
Java developer can easily complement clustering information with
profiling. Profiling is a key element allowing business users to recog-
nize a good clustering from a bad one.

12.4 Summary

This chapter demonstrated how to design and implement software to
solve practical business problems using data mining. We used JDM
and also JDBC since data manipulations such as filtering datasets or
computing some aggregate values are needed in such projects and
readily performed in SQL. We have shown how to mix data mining
operations with cost structures to improve business performance and
how the JDM TestMetrics can be used for this. And we have shown,
through the use of AttributeImportance and Clustering models, how to
help business users make greater use of business data.
We have taken examples from customer relationship management
(CRM) since the foundations of CRM processes, such as marketing