Lab 4: Viewing and Testing Predictions

In this lab you will learn about tools used to view and test predictions. The following are three tools used to determine if the Prediction model is yielding good predictions:

• Decision Trees: A Prediction model consists of a set of decision trees and a summary network called a dependency network. A dependency network summarizes the dependencies (predictive relationships) described by the decision trees.

  You use the Prediction Model Viewer to view both the dependency network and the individual decision trees. For more information about using decision trees, see Using Decision Tree View.

• Recommendation and Datafit Scores: You will learn about recommendations and Datafit scores in Lab 10: Automating Site Operations.

• Prediction Viewer: Using the Prediction Viewer tool, you can manually create sample cases and analyze prediction results. You can also determine a good setting for the popularity penalty. Popular items that are purchased frequently are considered very well known, and recommending them to a user is not required. A higher value for the popularity penalty reduces the likelihood that a popular item is included as a recommendation.

  The value for the Popularity Penalty ranges from 0.0 to less than 1.0. The default value is 0, which results in no penalty for popular items. This is the recommended value to maximize the sales of the most popular products. A setting of 0.5 to 0.7 is usually appropriate when users are most likely to be familiar with the most popular items.

  The Prediction Viewer tool is included in the Commerce Server 2000 Resource Kit.

Note

• The Prediction Viewer tool is different from the Prediction Model Viewer, which is a part of Commerce Server 2002.

Copyright © 2005 Microsoft Corporation.
All rights reserved.