After deciding which mining model is best, you need details to make predictions actionable. Microsoft enables a couple of ways to access those details.

SQL Server Integration Services (SSIS) includes the Data Mining Query transformation which uses input data with an existing mining model to return a prediction. You can automate predictions by including a Data Mining Query transformation in a SSIS package and returning predictions in a regular data flow.

To make the prediction exporting process easier to understand, I will use SQL Server Data Tools to manually export a batch of selected prediction details.

With a valid mining model open, click the Mining Model Prediction tab. To export a set of results you need to join a mining model with a set of input columns. In the Mining Model area, click Select Model and select the mining model you want to use for predictions. In the Select Input Table(s) area, choose a set of data to be used as inputs for predictions. Leveraging the SSAS Data Source View is the easiest way to ensure prediction inputs are compatible with the chosen mining model.

If mining model column names match input table column names, SQL Server Data Tools automatically connects Mining Model and Select Input Table(s) columns. If column names differ between the mining model and input table, you must connect the necessary columns manually to ensure accurate predictions.

Once the mining model and input table are properly connected, drag columns from the available sources or enter data elements to be exported. In my example, I chose a mix of player names from my input table and predictions from my mining model. The Prediction Function source includes many fields beyond Predict (i.e., the predicted value) and PredictProbability (i.e, the estimated probability of the predicted value). If you are predicting the likelihood of responding to a direct mail piece, you might wish to export names and addresses in addition to the prediction and probability so that you can mail your offering to only recipients predicted to respond.

Because the export uses the DMX language, you can further optimize your exported results. Right-click in the work area and choose Query to view and edit the generated DMX code. At the end of the prediction join query, you can add DMX commands like WHERE … to filter out results and ORDER BY … to arrange the results in a desired order. My example shown excludes any player with the probability of being an All Star less than 0.4 and orders results by highest likelihood to be an All Star.

If a small volume of data is to be exported, simply right-click the results, select Copy All and paste the data into Excel. For larger data volumes or if you need to join the results with other data, click the Save query resulticon to export the results directly into a database. Select an available database, enter a table name and click Save. As long as you have adequate permissions, the export process will create a simple table in the database and insert the resulting data into the table.

Once the results are exported from SSAS, the task of acting on those predictions is up to you.

Analysis Services supports much more complex data mining efforts than I demonstrated, but Microsoft's tools make accurate predictions manageable. I encourage you to try those tools with data and questions of your own.