Data Mining Process - Java Data Mining: Strategy, Standard, and Practice - page 77

Web

Call Center

Data Warehouse

Application Server

Mobile

Data Mining Engine

.
.
.

Campaign
Management

Operational
Data Stores

Applications

Figure 3-14 Enterprise software architecture involving in-database data mining.

Of course, the architecture just described is only one of the many
possible architectures. If an in-database data mining tool architecture
is used, as illustrated in Figure 3-14, either the data mining engine
and data mart are merged, or the data mining engine, data mart, and
warehouse components are merged to further simplify the architec-
ture. Some businesses have concerns over how the mining activity
may impact data warehouse performance. However, with newer
technologies [Oracle 2006], processors can be effectively partitioned
to separate mining from query and reporting demands on the data
warehouse. The need for a data mart as opposed to mining the data
warehouse directly may depend on the concerns of the database
administrator (DBA) for managing machines, disk space, and overall
system performance.
We have seen where data mining fits into an overall enterprise
architecture. But as a matter of process, how do we incorporate data
mining into an operational system?

3.5.2 Incorporating Data Mining into Business Operations

If we have followed the CRISP-DM process, we will have identified
the data needed for mining, determined the steps necessary for pre-
paring the data for mining, performed enough experimental mining
to know which models work well for the business problems at hand,