Data Mining - Professional SQL Server Analysis Services 2005 with MDX - page 761

model using the wizard and the remaining models within the same mining
structure using the data mining editor. The mining model wizard is self-ex-
planatory and you can explore the creation of other mining models such as
Microsoft Sequence Clustering, Microsoft Regression Trees, Neural Net-
works, and Sequence clustering using the data sets available in the Adven-
ture Works DW sample relational database.
OLAP Mining Models
Certain types of business problems necessitate the use of aggregated data
for analysis instead of individual input rows. For example, assume a customer
buys several products from various stores. You might want to segment the
customers not only by their individual attributes but also by the total amount
they have spent in the stores. The mining model would require aggregated
sales from the various purchases made by the customer, and include that
amount as an input attribute to the clustering mining model. You can certainly
add such a column to the customer table using a named query, but if the
sales information table has billions of records, the aggregation by the relation-
al data source will be slow. You should also consider maintainability of the
mining model because you might want to process the models on a periodic
basis. You indeed have a better solution than aggregating the data at the rela-
tional data source level. What better way to aggregate data than by creating a
cube?
Because Analysis Services helps you create cubes as well as mining models,
Analysis Services 2005 provides a way of creating mining models from
cubes. Such mining models created on top of a cube are called OLAP mining
models since the data source for the mining models is a cube and cubes con-
tain OLAP data. Analysis Services 2005 also provides you the functionality of
creating new cubes that include content from the created mining model along
with the original cube which provides you the power and flexibility to analyze
the cubes based on patterns discovered by the mining model. Such an ana-
lysis can further help you understand your data better and make better busi-
ness decisions. You will create and analyze cubes containing mining model
content in this section. You will use the AnalysisServices2005Tutorial you cre-
ated earlier to create OLAP Mining Models in this chapter. When you down-
load the samples for this topic you will find the AnalysisServices2005Tutorial