Data Mining Process - Java Data Mining: Strategy, Standard, and Practice - page 75

allow full interoperability among the objects produced either
through the GUI or API(s).
Tools can further be characterized along other dimensions: data
access options, DME location, and mining object storage . Historically,
data mining tools were file based, storing all mining objects and
metadata in flat files at the operating system level. With the preva-
lence of data stored in databases, nearly all tools support accessing
data in commercial relational databases. A DME that accesses remote
data exists as a standalone server, an operating system process or set
of processes that loads data from files or extracts data from the data-
base. Such a DME then mines the data and saves any results back to
the file system. This is illustrated in Figure 3-11(a). Other systems
leverage the relational database itself to store all mining objects and
metadata, thereby keeping the mining process under database con-
trol, as illustrated in Figure 3-11(b).
Other vendors take a database-centric approach where the
algorithms are moved to the data, instead of the data to the
algorithms, as illustrated in Figure 3-12. As data volumes continue
to grow, the time and space required to move data outside the data-
base also grow, sometimes dwarfing the data mining execution time.
In-database mining also avoids the management of extracted data,
which includes disk space, memory, and security.
With this foundation in place, we now take a look at an enter-
prise architecture involving data mining. In Figure 3-13, we see sev-
eral operational data stores (ODSs), containing perhaps customer,
sales, service, account, etc., that support various applications, such
as online retailing, call centers, business intelligence dashboards,
business activity monitoring (BAM), mobile customers, campaign

Data Mining Tool Data Mining Tool

Data Mining Engine Data Mining Engine

File
System
(MOR)

File
System
(Data Source)

File
System
(Data Source)

Database
(Data Source)

Database
(MOR)

Database
(Data Source)

(a) (b)

Figure 3-11 Data mining tool data access and mining object storage architectures.