Chapter 10. Techniques for data mining in an operational warehouse 389
Customer table to identify the relative RISK level of a particular customer. The
SQL for this query is shown in Example 10-5.
Example 10-5 Joining scoring output with other data warehouse data
SELECT C.CUSTOMER_ID, C.RISK, L.LOAN_ID, L.LOAN_AMOUNT
FROM CUSTOMERS ANALYZE_TABLE (IMPLEMENTATION ‘PROVIDER=SAS;
ROUTINE_SOURCE_TABLE=USER1.SOURCE_TABLE;
ROUTINE_SOURCE_NAME=FRAUD;’) AS C,
LOANS AS L,
WHERE C.CUSOMER_ID = L.CUSTOMER_ID
AND C.RISK > 50
ORDER BY C.RISK DESC;
This SQL query returns an ordered list (by RISK from highest to lowest) of
customers. Each record will list the customer, the risk score, the loan account
number and the loan amount. The business analyst or loan officer can then take
action on the high risk customers as is appropriate per business policy.
10.4 Deploy and visualize mining results with IBM
Cognos 10 Business Intelligence
Regardless of the data mining technique used in the operational data
warehouse, the data mining and scoring results must be presented to and
visualized by users in some way to have business value. The results can be
leveraged in a variety of ways including:
Ad hoc visualization of the mining or scoring results.
Scoring results can be used to augment dimensional or fact table records.
Scoring results can be used to drive or trigger business processes or to
influence decisions in business processes.
Visualization can take many forms, and we have seen examples of visualizations
in the previous sections. We saw examples of the built-in visualizations in the
InfoSphere Warehouse 10.1 Design Studio that are customized for each data
mining method. We also saw examples of integration of scoring results using
InfoSphere Warehouse and SAS using standard SQL techniques.
IBM offers IBM Cognos 10 Business Intelligence to provide the most robust end
user and business analyst visualization and interaction with the data mining
results. The Customer Insight InfoSphere Warehouse Model Pack includes a
practical example of using Cognos to use data mining results. The Customer
Insight Pack contains an example of customer segmentation using the clustering