Database Reference
In-Depth Information
Association Rules
We have already explained that the association model for the Northwind
case study requires two views, namely,
AssocOrders
and
AssocLineItems
.Once
these views are prepared, we can set the model parameters like support and
confidence (called
probability
in Analysis Services) and deploy the model. In
our case, we just show the results to give the look and feel of the tool and
discuss the DMX commands since the number of records is still low to produce
meaningful results. Figure
9.11
shows the association rules obtained, along
with their probability. Figure
9.12
shows some of the itemsets obtained. For
example, we can see that the products
Sirop d
'
erable
and
Sir Rodney
'
sScones
have a support of 55, which means they appear together 55 times in the
database. The reader can also compare the itemsets and the corresponding
rules that those itemsets yield. For example, the itemset (
Sirop d
'
erable, Sir
Rodney
'
sScones
) produces two rules, since the order of purchasing does not
matter:
Sirop d
'
erable
→
Sir Rodney
'
sScones
,and
Sir Rodney
'
sScones
→
Sirop
d
'
erable
.
Fig. 9.11
Association rules for the Northwind case study
There are two common uses of an association model: to discover infor-
mation about frequent itemsets and to extract details about particular rules
and itemsets. For example, we can retrieve a list of rules that were scored
as being especially interesting or create a list of the most common itemsets.
This information can be obtained through a DMX content query or browsing
this information by using the Microsoft Association Viewer. The next content
query returns details about the parameter values that were used when the
model was created:
SELECT MINING PARAMETERS
FROM $system.DMSCHEMA MINING MODELS
WHERE MODEL NAME =
'
Association
'
Search WWH ::
Custom Search