1297
Chapter 57: Data Mining with Analysis Services
57
■ (^) Some dimension attributes based on numeric or date data may appear to the data
mining interface with a text data type. This is because data mining uses the Name
column’s data type instead of the Key column. If this causes an issue, remove the
Name column property from the dimension attribute, or add the same column to
the dimension a second time without using the Name column property.
■ The portion of cube data to use for training is defi ned via the mining structure’s
cube slice.
■ A Lift Chart cannot be run against cube test data, so model evaluation requires
either test data in a relational table, or some strategy that does not rely on the
tools of the Mining Accuracy Chart.
Using a cube as a mining data source can be effective, providing access to what is often
large quantities of data for training and testing, and providing the ability to create a
dimension or even an entire cube based on the trained model.
Summary
Data mining provides insights into data well beyond those provided by reporting, and
Analysis Services streamlines the mining process. Although the data must still be prepared,
mining models hide the statistical and algorithmic details of data mining, enabling focus
on analysis and interpretation.Beyond one-time insights, you can use trained models in applications to allocate scarce
resources, forecast trends, and identify suspect data, and for a variety of other uses.c57.indd 1297c57.indd 1297 7/31/2012 10:35:04 AM7/31/2012 10:35:04 AM
http://www.it-ebooks.info