Once this information begins flowing through to the warehouse, the business decision makers can now make essential data mining predictions based on this missing “piece of the pie”. Subsequently, this new information may lead to the realization that other pieces of information are missing, or have errors. As a result, the modification of the OLTP applications and the warehouse begin again.
Figure 1.1
Closed Loop Data Mining Process — Hardware Sales Example
Operational Data Mining
The Operational Data Mining methodology positions data mining as an integral component in an application’s architecture. This type of data mining allows for real-time application access to a data mining model for the purpose of decision making. Not to be confused with an integrated n-tier application component, data mining in an operational data mining context, will serve as a resource for decision making requirements, as related to the sustenance of a dedicated business process. In other words, data mining can serve as means to add decision making capability to a business process. Similar to the closed loop methodology, operational data mining is built upon the information gathered from typical OLTP applications and data warehousing and OLAP sources. However, during the course of OLTP processes, the application(s) make direct reference to data mining decision trees in order to continue along logical processing paths.
This type of data mining requires the extensive refinement of data mining through a comprehensive closed loop process. Once the series of data mining models defined and refined in the closed loop process are reliable and produce consistent results, they can then serve as a component of an enterprise decision making process integrated within an application or business process. The following example details how the same hardware company makes credit approval decisions for users wishing to finance their purchases over a period of time.
Hardware Company Credit Approval Example
The hardware company’s credit organization can use data mining in order to expedite and execute the approval process. The facilitation of credit applications can be integrated into the existing e-commerce storefront in the form of an online application or an integrated a call center to handle the application calls.
The credit organization can take advantage of previously built and refined data mining models, established by a pre-existing closed loop data mining process, to enable quick and reliable approval or denial decisions for credit applicants. The application will use the credit applicant’s application information by comparing it against a data mining model.
The data mining model(s) in this example can be used not only to determine if the applicant has bad credit history and subsequently deny the application to finance the hardware, but if the credit application is approved (by comparing application information against an approval model), a separate mining model can be used to determine other add-on products to target this applicant with.
Figure 1.2
Operational Data Mining Process — Hardware Credit Approval Application Example