Web Services - Java Data Mining: Strategy, Standard, and Practice - page 275

mining to many production applications. But the current process is
not scalable for its needs because data analysts must maintain hun-
dreds of models and produce results manually. Moreover, the IT
group needs to integrate data mining results manually by deploy-
ing to the appropriate applications. In addition, ABCBank wants to
be able to change DMEs used in the production environment and
to have the ability to integrate with multiple DMEs. Due to legacy,
their enterprise applications are written using multiple computer
languages such as Java, C++, C, and COBOL.
From the perspective of ABCBank's business analysts, automat-
ing data mining processes and embedding the results with produc-
tion applications will produce more fruitful results than the current
limited integration. Embedding data mining results in the produc-
tion applications will enable business executives of all levels to
make more informed decisions using the latest data. In addition,
ABCBank call-center representatives can interact with customers
more effectively if the customer attrition likelihood, profitability,
and appropriate cross-sell products are displayed along with the
customer profile.
From the perspective of ABCBank's data analysts, to ensure the
quality of the data mining results they would like to maintain con-
trol of the specification of model build settings such as algorithm
and function level settings and the attributes to be used; they
would also like to be able to add or remove attributes. Once the
mining attributes and model build settings are specified for a
specific problem, model quality can be maintained if the model is
periodically updated with the new data, thereby learning new pat-
terns. In the future, analysts would like to build a mining solution
once for a problem and deploy the required settings to production
applications. The automated processes should be able to rebuild the
mining models with new data at the specified intervals. Over time,
if the models do not meet quality thresholds, DM Lab must be
notified to improve them. Automation enables the data analysts to
manage many models easily, and allows them to focus on improv-
ing non-performing models and addressing new business prob-
lems.
From the perspective of ABCBank's IT department, the data
mining and application integration architecture must be flexible
enough to support both current and future applications. The IT
department also must be able to run computationally intensive data
mining operations in a separate machine without affecting produc-
tion applications. To have the flexibility to change DMEs to avoid