Chapter 9 Exercise 3

3. Find a commercial data mining tool and outline the steps needed to perform a market basket analysis.

Answer is variable. Oracle has a Java-based tool, and the steps for this analysis are described in the Workbook. More specialized software exists (SAS and megaputer). See the documentation for the details on the setup. Once the data is configured correctly, running the analysis is often a simple matter of issuing a command or selecting a menu option. The biggest challenge often lies in setting up the data correctly. SQL Server and WEKA are probably the easiest.

SQL Server uses transaction data so it is the easiest. The basic steps are:
� Create a new BI project.
� Define the data source/connection.
� Define the data set/query data.
� Create a new BI association model and select the data, possibly write a query.
� Build the model and deploy it, then browse it.
� Set the cut-off parameters and run the model again.

WEKA, wants transaction data stored on a single row, so the biggest headache is getting the data organized. From there, WEKA is basically menu driven, so connect to the file, choose the variables/model, run it, and examine the results. But, configuring the data is a challenge; unless perhaps WEKA fixes this.
Most of the following questions require an OLAP cube processor. You should have access to SQL, an OLAP browser within the DBMS, or a PivotTable. For the data mining tools, if you do not have access to specialized software, you can use Excel for simple analyses.