Databases Reference
In-Depth Information
Itemset key Definition
The Itemset is the unit of work, the unit of business activity. To discuss
what Objects are in, or not in, an Itemset the Itemset must have the iden-
tity properties that allow each Itemset to be uniquely and consistently
identified such that each Itemset will always be distinct from, and never
confused with, all other Itemsets. In other words, each Itemset must have
a key unique to itself. The source system that is the system of record for the
Itemset transactions may, or may not, have within itself a feasible key that
can uniquely identify each and every Itemset. If that source system has a
key that maintains the identity properties of the Itemset and performs well
in the relational platform of the analysis, then that source system transac-
tion key may be defined as the Itemset key. If the source system does not
have a key that simultaneously maintains the identity properties of the
Itemset and performs well in the relational platform of the analysis, then
the ETL application (ETL will be discussed in Chapter 7) that presents
data to the Market Basket Analysis application will generate a key that
does maintain the identity properties of the Itemset and performs well in
the relational platform of the analysis.
For the examples listed later in this chapter, the Itemset key will be
defined as an integer data type. This should not be interpreted to mean
that an Itemset key must be an integer data type. Instead, this use of inte-
gers simply means that integers make acceptable sample data.
object key Definition
The Driver Object and Correlation Object are interchangeable. Both
objects should therefore share the same key structure. The Market Basket
Analysis application receives its data from a data warehouse. Assuming
the data warehouse already has a d efined key structure for the objects
in the Itemset, the Market Basket Analysis application should be able to
leverage those keys already in the data warehouse.
If the data warehouse does not have a d efined key structure for the
objects in the Itemset, then the Market Basket Analysis application will
not be able to identify occurrences of the individual objects in an Itemset.
If that is the case, then the data warehouse is not able to support a Market
Basket Analysis application. The entity definitions and key structures of
the data warehouse must be remediated before the Market Basket Analysis
application can leverage data from the data warehouse.
