With this result, we proceed with the mailing and await the actual
customer responses. It will be important to compare actual responses
with those predicted to determine how accurate the model was in
practice. This type of feedback is key to determine if the introduction
of data mining indeed met the business objectives.
As part of deployment, the data miner produces a report summa-
rizing the steps necessary to introduce data mining to the campaign
process and the details of building and evaluating the models.
This chapter introduced the JDM API using a response modeling
business problem. We followed the phases of the CRISP-DM method-
ology to illustrate where the API applies, putting the use of the API
in the context of not only solving a data mining problem, but leverag-
ing its results. We saw how the notion of lift could be used to assess
the quality of a model and how it can impact the selection of a final
model or strategy. There are other techniques for assessing model
quality, such as a confusion matrix and receiver operating characteristics
(ROC), which will be introduced in Chapter 7.
[JDBC 2006] http://java.sun.com/products/jdbc.
[SQL 2003] ISO/IEC 9075-2:2003, Information technology-Database lan-
guage-SQL-Part 2: Foundation (SQL/Foundation) , International Stan-
dards Organization, 2003.
[SQLJ 2003] ISO/IEC 9075-2:2003, Information technology-Database
languages-SQL-Part 10: Object Language Bindings (SQL/OLB) , Inter-
national Standards Organization, 2003.