Example 3-5 shows data from a view of a sales table. It includes sales for three of the many products sold in four regions. This data is not suitable for mining at the product level because sales for each case (product), is stored in several rows.
Example 3-6 shows how this data could be transformed for mining. The case ID column would be PRODUCT
. SALES_PER_REGION
, a nested column of type DM_NESTED_NUMERICALS
, would be a data attribute. This table is suitable for mining at the product case level, because the information for each case is stored in a single row.
Oracle Data Mining treats each nested row as a separate model attribute, as shown in Example 3-7. (Note that the presentation in this example is conceptual only. The data is not actually pivoted before being processed.)
Example 3-5 Product Sales per Region in Multi-Record Case Format
PRODUCT REGION SALES ------- -------- ---------- Prod1 NE 556432 Prod2 NE 670155 Prod3 NE 3111 . . Prod1 NW 90887 Prod2 NW 100999 Prod3 NW 750437 . . Prod1 SE 82153 Prod2 SE 57322 Prod3 SE 28938 . . Prod1 SW 3297551 Prod2 SW 4972019 Prod3 SW 884923 . .
Example 3-6 Product Sales per Region in Single-Record Case Format
PRODUCT SALES_PER_REGION (ATTRIBUTE_NAME, VALUE) ------ -------------------------- Prod1 ('NE' , 556432) ('NW' , 90887) ('SE' , 82153) ('SW' , 3297551) Prod2 ('NE' , 670155) ('NW' , 100999) ('SE' , 57322) ('SW' , 4972019) Prod3 ('NE' , 3111) ('NW' , 750437) ('SE' , 28938) ('SW' , 884923) . .
Example 3-7 Model Attributes Derived From SALES_PER_REGION
PRODUCT SALES_PER_REGION.NE SALES_PER_REGION.NW SALES_PER_REGION.SE SALES_PER_REGION.SW ------- ------------------ ------------------- ------------------ ------------------- Prod1 556432 90887 82153 3297551 Prod2 670155 100999 57322 4972019 Prod3 3111 750437 28938 884923 . .