Specifying mining data

Use the following command to run the sample script bankingCreateMiningData.db2 to create the DM_MiningData values and to insert them into the table IDMMX.MiningData:

db2 -stf bankingCreateMiningData.db2

The first part of the sample script bankingCreateMiningData.db2 creates a DM_MiningData value and inserts it into the table IDMMX.MiningData. The column name and type information is extracted from DB2® catalog tables:

INSERT INTO IDMMX."MININGDATA"("ID","MININGDATA")VALUES (
   'AliasEqualToColumn',
   IDMMX.DM_MiningData()..DM_defMiningData('BANKCUSTOMERS'));

The next part of the sample script defines column aliases that differ from the column names in the input data source using method DM_setFldAlias.

INSERT INTO IDMMX."MININGDATA"("ID","MININGDATA")VALUES (
   'AliasDifferentToColumn',
   IDMMX.DM_MiningData()..DM_defMiningData('BANKCUSTOMERS')
   ..DM_setFldAlias('CLIENT_ID','CUSTNO')
   ..DM_setFldAlias('PROFESSION','JOB')
   ..DM_setFldAlias('NBR_YEARS_CLI','LOYALTY'));

The next statement creates a view to use only a sample of customers as input data. This example defines a repeatable random sample of 10 percent. The sample must be repeatable to support multiple passes during model training:

CREATE VIEW "CUSTSAMPLE"
   AS SELECT * FROM "BANKCUSTOMERS" TABLESAMPLE SYSTEM(10) 
                                    REPEATABLE(3);
   
INSERT INTO IDMMX."MININGDATA" VALUES (
   'CustomerSample',
   IDMMX.DM_MiningData()..DM_defMiningData('CUSTSAMPLE'));

The following statement uses a predicate (WHERE clause) to specify the list of input records:

INSERT INTO IDMMX."MININGDATA" VALUES (
   'MarriedCustomers',
   IDMMX.DM_MiningData()..DM_defMiningData('CUSTSAMPLE')
   ..DM_setWhereClause('MARITAL_STATUS=''married'''));