Lab 2: Viewing and Interpreting Analysis Models

In the lab, you use Commerce Server Manager and the Business Desk application to view and interpret Prediction and Segment models.

This lab contains the following steps:

  • Step 1: View the Prediction model using Commerce Server Manager
  • Step 2: View the Segment model using Business Desk

Step 1: View the Prediction model using Commerce Server Manager

  1. Click Start, point to Programs, point to Commerce Server 2002, and then click Commerce Server Manager.

  2. In Commerce Server Manager, expand Commerce Server Manager, expand Global Resources, expand Predictor on <server name>, expand Predictor Service, and then select Models.

  3. In the Commerce Server Manager content pane, right-click Software Predictions, and then select View Model.

  4. In the Software Predictions screen, you can see the dependency network between your products.

    Software Prediction Dependency Network

    For more information about viewing the dependency network, see Using Dependency Network View.

  5. In the Software Predictions screen, click Find on the toolbar.

  6. In the Node Finder dialog box, select NFL Fever 2000.QTY from the list, and then click Go to Node. Note the following:

    • The selected product is represented as a green node, and the other products it predicts are represented as purple nodes.

      Dependency Network for NFL Fever

    In this example, the Prediction model shows a positive correlation between NFL Fever 2000 and NBA Inside Drive. If a user is interested in one of the two products, it is very likely that the user will be interested in the other product as well. Two or more products may have a positive or negative correlation with each other. In a negative correlation, a user who chooses one of the products is unlikely to choose the other. For example, a negative correlation exists between audio CDs and audiocassettes. You use the decision tree view to determine whether the relationship between products is positive or negative.

  7. In the Software Predictions screen, click any white space to clear the NFL Fever 2000.Q node, and then click Find on the toolbar.

  8. Using the Node Finder, you can find any node in the model. In the Node Finder dialog box, select Mastering Web Application Development Using Visual InterDev 6.QTY from the list, and then click Go to Node.

  9. In the Software Prediction Model screen, double-click the node Mastering Web Application Development Using Visual InterDev 6.QTY.

  10. In the Decision Tree forMastering Web Application Development Using Visual InterDev 6.QTY screen, note the following:

    • The Decision tree for Mastering Web Application Development Using Visual InterDev 6.QTY contains a set of IF-THEN rules that are triggered by a user's purchasing behavior.

    • Each path in the tree (directed by an arrow) corresponds to an IF<Condition> THEN<Prediction> rule.

    • A blue oval node represents a property in the conditional part of an IF-THEN rule, and a square node contains the prediction part of the rule.

      Decision Tree for Mastering Web Applications

    In this example, Microsoft Site Server 3.0 is the conditional product with two conditions: missing and not missing. The numbers in parentheses represent the number of times that that specific condition was true.

  11. In the Decision Tree for Mastering Web Application Development Using Visual InterDev 6.QTY screen, double-click the square node located in the upper-right corner labeled, “missing”.

  12. In the Details forMastering Web Application Development Using Visual InterDev 6.QTY dialog box, note the following:

    IF-THEN rules for Mastering Web Applications

    • IF-THEN condition in detail.
    • The IF box displays the condition, and the THEN box displays the probable outcomes of buying 0,1,2, or 3 copies of the product.

    In this example, if the user has not purchased a copy of Site Server 3.0, then the probability of the user not purchasing Mastering Web Applications for Development Using Visual InterDev 6.0 is 98 percent.

  13. Close the Details forMastering Web Application Development Using Visual InterDev 6.QTY dialog box.

  14. In the Decision Tree for Mastering Web Application Development Using Visual InterDev 6.QTY screen, click the back arrow on the upper left corner.

  15. In the Software Predictions screen, click any white space to clear the Mastering Web Application Development Using Visual InterDev 6.QTY node, and then click Find from the toolbar.

  16. In the Node Finder dialog box, select Visual Studio 6.0 Professional Edition from the list, and then click Go to Node.

  17. In the Software Predictions screen, double-click the node Visual Studio 6.0 Professional Edition.

  18. In the Decision Tree for Visual Studio 6.0 Professional Edition.QTY screen, note the following:

    Decision Tree for Visual Studio 6.0

    • The screen shows the decisions tree for Visual Studio 6.0.
    • Visual Studio 6.0 Professional Edition has a predictive dependency on MSDN Professional Subscription, Microsoft Windows 2000 Server, and Microsoft Visual J++ 6.0 Professional Edition.
  19. In the Visual Studio 6.0 Professional Edition.QTY screen, double-click the square node labeled missing (982), which connects to MSDN Professional.

    Ee799115.note(en-US,CS.20).gifNote

    • The number of users that satisfy the IF condition are presented in brackets ''()''. In this example, there are 982 users who belong to the node labeled missing (982).
  20. In the Details forVisual Studio 6.0 Professional Edition.QTY dialog box, note the following:

    IF-THEN rules for Visual Studio 6.0

    Of the users who did not purchase Windows 2000 Server and MSDN – Professional Subscription, 98.7 percent did not purchase Visual Studio 6.0 Professional Edition either.

  21. Close the Details forVisual Studio 6.0 Professional Edition.QTY dialog box.

  22. In the Visual Studio 6.0 Professional Edition.QTY screen, double-click the bottom-most square node labeled not missing (31) that connects to MSDN Professional.

  23. In the Details forVisual Studio 6.0 Professional Edition.QTY dialog box, note the following:

    IF-THEN rules for Visual Studio 6.0

    • 8.6percent of users who purchased both Windows 2000 Server and MSDN – Professional Subscription, did not purchase Visual Studio 6.0 Professional Edition.
    • 20 percent of the users purchased more then one version.
  24. Close the dialog box, the Decision Tree for Visual Studio 6.0 Professional Edition.QTY screen, and Commerce Server Manager.

Step 2: View the segment model using Business Desk

  1. Start the Predictor Demo Business Desk.

  2. In Analysis, click Segment Viewer.

  3. In the Segment Viewer screen, select Software Segments, and then click Open.

  4. In the Segment Viewer: Software Segments screen, note the following:

    Software Segments Model

    • The segment hierarchy in the navigation box represents the entire population. (100 percent). The Segment hierarchy consists of segments and segment groups. A segment is a collection of users with similar property values, for example, users who have made similar purchases. A segment group is a set of similar segments.
    • The Summary of Population box displays a list of the most likely properties and their values in the entire population. The longer the red bar, the more likely the Property/Value pair is in the population.
  5. In the Segment hierarchy pane, expand Population (100%), expand all the segments and segment groups, and then note the following:

    Segment 5

    • The summary of Segment 5 displays the most likely property/value pairs associated with users in this segment. In this example, Money 2000 Business & Personal.QTY with the value 3 indicates that most users in this segment purchased 3 copies of Microsoft Money 2000 Business and Personal.
    • You can compare segments: The Compare Segment 5 to dropdown box has Population selected by default. However, you can select any segment or segment group for a comparison between the selected segment and rest of the segments present in the segment model.
    • Yellow bars indicate the property/value pairs that increase the membership a user in cluster 5. In this example, Baseball 2000.QTY with value of 1 means a user who purchased a copy of Microsoft Baseball 2000 becomes a member of segment 5.
    • Blue bars indicate the property/value pairs that decreases the membership of a user from segment 5. In this example, Windows 98 Second Edition.QTY with a value of 2 means a user who bought two copies of Microsoft Windows 98 Second Edition is less likely to be a member of segment 5.
    • Purchases of Baseball 2000 and Microsoft NBA Inside Drive 2000 place a user in segment 5. This implies that the users in cluster 5 are sports computer games fans compared to the rest of the segments in the total population.
    • You can rename segments to make the names meaningful and easier to browse.
  6. In Segment hierarchy, select Segment 5, and then click Rename.

  7. In the Rename Segment dialog box, in the Rename to box, type Sports Computer Games Fans, and then click OK.

  8. In Segment hierarchy, select Segment 7, and then note the following:

    • Users in the cluster 7 purchased Site Server 3.0, Commerce Edition and SQL Server 7. This implies that the users in cluster 7 are Web Developers.
  9. In Segment hierarchy, select Segment 7, and then click Rename.

  10. In the Rename Segment dialog box, in the Rename to box, type Web Developers, and then click OK.

  11. In Segment hierarchy, select Segment Group 14, and then note the following:

    • The Comparing Segments section provides summary and other comparison information for segments as well as segment groups.
    • You should view all the segments and segment groups within a model to get a better understanding of the segment model.
  12. Analyze all segments in Segment Group 14 to deduce that it represents home users.

  13. Rename Segment Group 14 to Home Users.

  14. In the Segment hierarchy pane, select Segment Group 12, and then note that the group represents small business users.

  15. Rename Segment Group 12 to Small Business Users.

  16. In the Segment hierarchy pane select Segment 3, and then note that the group represents developers.

  17. Rename Segment 3 to Developers.

    Segment Viewer

  18. In the Segment Viewer: Software Segments screen, click Report on the toolbar to view the summary report of the entire Segment model.

    • You can save this report to a file by saving the Model Summary Web page from your Internet browser.
    • You can also export user lists. User lists are used by Commerce Server 2002 to create a direct mail campaign to target users who are likely to buy a specific product.
  19. In the Segment hierarchy section, select Web Developers, and then click Export on the toolbar.

  20. In the Export Segment to List dialog box, do the following:

    Use this To do this
    Enter list name Type Web Developers.
    Minimum probability Verify that the value is 0.50.

    In Commerce Server 2002, every user is a partial member of all the segments. Segments tend to overlap with less distinct boundaries. So, a user may be 80% member of one group and a fractional member of the rest. The higher the fractional value assigned to the user, the closer the property values of the user are to the property values of the typical user in the segment. This fractional value of a user is the probability of membership for the given segment.

    The Minimum probability sets a threshold criterion for including a user in the export list. If the probability of membership of a user exceeds the specified threshold, the user is not included in the export list.

    If the Minimum probability is set to be high (close to 1), then only a small number of users will be included in the user list. However, if the value is set to be low (close to 0), then many users will be included in the export list.

  21. Click OK.

Copyright © 2005 Microsoft Corporation.
All rights reserved.