September MDX Puzzle Solution Revealed - 21 Sep 2000

The FoodMart 2000 database includes a virtual cube called Trained Cube, with a data-mined dimension called Customer Pattern. This dimension is based on a decision-tree model called Customer Pattern Discovery. To discover how well this model predicts a particular member card, write an MDX query that returns the percentage of customers identified by the [Customer Pattern].[All].[Customers.Name.Yearly Income = $10K - $30K] dimension member that have a Normal Member Card. (Hint: The answer will be greater than 90 percent.)

The query in Listing A uses the WHERE clause to highlight the customers who have a yearly income of $10,000 to $30,000. The NonEmptyCust set contains all customers within this income bracket. The calculated measure [Normal] divides the total number of customers in this range by the number of people in that group who have a Normal Member Card. The query formats the resulting value as a percentage and displays it as a column in the report.