Summary

• Chapter 5 introduced the fictional retailer SavingsMart and used Analysis Services to build a mining model that predicted days between shipments and quantity. This chapter expands on the results obtained from the mining model to create a new shipment method.

• Communicating with Analysis Server is accomplished using the Microsoft Active Data Objects Multidimensional (ADOMD) library. It also involves the installation of XML for Analysis specification 1.1 and the latest version of the MSXML parser.

• A modified version of the LoadSampleData program used in Chapter 5 was used to apply prediction results and generate new purchases for 2002. Using database files available for download from the book's Web site, the reader simulated results with the new shipment method. Initial observations indicate that although the total shipments will be reduced, customers will be less likely to find the products they need.

• Since mining-model results are highly dependent on the training datasets utilized, an alternative database file is provided. Reprocessing the mining model using this new database file results in more branches and the inclusion of an additional factor. After utilizing the new predictions, total shipments are reduced even further, as are missed sales opportunities.

• Closed loop processing involves the continual reprocessing of mining models and the reapplication of their predictions. This ensures that mining-model predictions remain as accurate as possible when conditions change. The LoadSampleData program provides an example of how this can be done.