Databases Reference
In-Depth Information
Itemset
In Chapter 5, the “Definition of an Itemset” section explained that every
enterprise has moments when the business performs its business func-
tions. Unfortunately, there is no universal Transaction Key. Unlike the
Entity Key, which is based on the data modeling concept of an entity,
and the Instance Key, which is based on individual iterations of an
entity, there is no data modeling concept that encompasses the indi-
vidual transactions of a restaurant, retail store, and service provider.
For that reason, the label Itemset cannot be replaced by something else
that is universally recognized as the unique identifier for transactions
or Itemsets.
That being the case, the definition of an Itemset is unique to each
enterprise. For a Fact table to represent the Itemsets of the enterprise to
a Market Basket Analysis application, that Fact table must also be able to
represent each individual transaction so that each individual transaction
can be identified separately from all other transactions. The first Market
Basket ETL function will extract data from a Fact table and then load that
data into a Ma rket Basket Table. Within the Market Basket Table, each
individual Itemset may be identified by a compound key composed of var-
ious entities, such as a date, time, place, station, and person, or by a simple
unique sequential key.
object
The Objects of a Market Basket Analysis application are also the Objects of
a Time Variant Market Basket Analysis application. The difference is time
variance, which means that Objects in a Market Basket Analysis applica-
tion are represented simultaneously by their Type 1 E ntity Key and by
their Type 2 Instance Key in a single Fact table row.
Chapter 7 ( Market Basket ETL) presented two hierarchical levels of
object. At the lower hierarchical level, the Object_Key represented an
individual object. At the higher hierarchical level, the Hierarchy_Key rep-
resented a grouping or class of objects. Both Object_Keys and Hierarchy_
Keys were incorporated into the Market Basket Analysis application. The
concept of juxtaposing an object and a class of objects was presented for
two reasons. The first reason is that such a juxtaposition is a value-adding
practice within Market Basket Analysis. The second reason is that it pro-
vides the conceptual foundation for Time Variant Market Basket Analysis.
