Case #3

Create a match when one event follows another within a specified time period. For example, identify cases where an account performs a withdrawal that is preceded by a change of registration information within seven days.

The recommended approach to this issue is to identify instances of the behavior of interest by creating a dataset that joins the tables involved. For example:

SELECT a.BO_TRXN_SEQ_ID, a.ACCT_INTRL_ID
From BO_TRXN a, CHG_LOG_SMRY b

Where:

a.ACCT_INTRL_ID = b.ACCT_INTRL_ID And a.DBT_CDT_CD = ‘C’
And b.ACCT_RGSTN_CHG_CT > 0
And b.CHG_DT  a.EXCTN_DT
And b.CHG_DT  (a.EXCTN_DT – 7)

Bring in additional tables (such as KDD_DATE_CONTROL) to control the lookback period.

An alternate strategy is to create a dataset that selects from the transaction table, and use an exists clause to detect the presence of an account registration change in the change log summary table.

Generate the matches by using a one-row rule pattern that consists only of a primary rule. Apply focus clustering to generate only one match for each account that exhibits this behavior. The pattern matches transaction building blocks because this field is specified as the dataset key. Use augmentation to add the building blocks for the relevant change log summary records.

Generally, it is more efficient to solve a business problem at the dataset-level, rather than through a sequence pattern, if the scenario involves a situation that consists of only two events or the behavior involves a series of events that are not time ordered.