Viewing with Dependency Network Browser
Dependency Network Browser views data mining content for decision tree mining models from a vantage point different from that of Data Mining Model Browser. Whereas Data Mining Model Browser allows you to view relationship and distribution information from the viewpoint of a single attribute, Dependency Network Browser gives you the ability to view the data mining model from the viewpoint of all attributes by relationship information alone, providing a wider perspective on the entire data mining model.
Consider the following example. You have constructed a data mining model based on product purchases for customers using store credit cards. The customer demographic information and product order information is quite rich; numerous input and predictable columns have been supplied for the data mining model. Using Dependency Network Browser, you can get an immediate sense of how well customer demographic information predicts other customer demographic information, for example, and how the same customer demographic information predicts product order information. Dependency Network Browser shows the relationships contained in every tree in the data mining model, information that would be difficult to extrapolate by viewing each tree in isolation.
After an attribute is isolated, Data Mining Model Browser allows you to view the details and distribution information for the relationships of the selected attribute.
Dependency Network Browser displays all of the attributes in the data mining model as nodes, with arrows indicating prediction links between nodes. For example, an arrow from an Exercise node to a Heart Rate node indicates that the Exercise attribute predicts the Heart Rate attribute. The strength of the prediction link can be viewed by the use of a slider on the left side of the window. If the slider is set to All links, all prediction links are shown, no matter how little they affect the related nodes. If the slider is set to Strongest links, only the strongest prediction links are displayed.
The nodes are color-coded to represent the selected node and the direction of predictability of related nodes. The nodes can also be moved by clicking and dragging to improve the view of relationships, or the Improve Layout button can be used to automatically distribute and resize nodes for better viewing.