Reporting Services makes an excellent delivery vehicle for your data mining results. In addition to exposing query results, you can also use features of Reporting Services to enhance the functionality of DMX. For example, DMX lacks GROUP BY functionality, so you can take advantage of the Reporting Services grouping ability in data mining reports.
Using this ability we can easily emulate the Classification Matrix functionality of the BI Dev Studio in a report that we can deliver to anyone. The first thing you will need is a mining model that predicts a discrete attribute and some test data. For this example I will be using a decision tree model build on the movie customer data available on this site to show a classification matrix (also known as a confusion matrix) for the Home Ownership column.
The first thing to do is to create a new Reporting Services project and connect to your Analysis Services database containing your model. When you generate your query, you will select the actual column from the test data set and the predicted column from the model - it's a good idea to give that column a name, e.g. Home Ownership, and Predicted Home Ownership, as shown below.
When you set up the grouping levels for your report, put the actual value (e.g. Home Ownership) into the group box and the predicted value (e.g. Predicted Home Ownership) into the details box. In the end, we want both items in the group, but RS requires that you put something in the details in the Wizard, so we will move this later.
After you finish the report wizard you need to edit the report layout to move the predicted home ownership from the details line to the grouping line as shown below. Then you can delete the details row from the report.
Next, right-click the grouping icon to the left of the row and select "Edit Group." Add Predicted Home Ownership to the list of grouping fields and click OK.
Finally you need to add a new column to the report. For the column header you can type "Count" and for the cell value enter "=CountRows()". Your final layout should look something like this (with some additional formatting):
Preview your report and you will recieve a thin-client, distributable, portable, classification matrix like this!