Decision Tree Model Viewer

This selection displays a successfully built decision tree model.

Exactly what information is displayed depends on how the viewer is invoked:

Menus

The following menus appear when the model is displayed in a new window:

Tabs

The viewer has the following tabs:

Tree

To see the rules, click the Tree tab. This tab is displayed by default when you double-click a tree model or the build result for a Decision Tree mining activity.

The rules are displayed in the bottom pane. If the rule is a constant, such as true, no rule appears in the bottom pane.

The default view shows all nodes and the attribute values used to determine splits. You can highlight a node to show the rule for a record to be included in that node.

To display leaf nodes only, click the Show Leaves Only checkbox. The leaf nodes or terminal nodes (nodes that are not split) are the nodes used to make predictions when the model is applied to new data.

The following options are available when Show Leaves Only is not checked:

To save the rules to a file, click the Save As icon (a yellow arrow above a graphic of a page) above the Nodes grid.

The following is displayed for each node in the Nodes grid:

You can do the following:

Examine Split Rules for Node

To examine the split rule for a node, click the Node ID.

The rule for the node is displayed in the bottom pane; the bottom pane has two tabs:

The following information is displayed on the Predicate tab in the lower pane for the selected node:

To make them more readable when exported to Microsoft Excel, the rules displayed in the Split Rules box are formatted with carriage control characters, for example,


CUST_INCOME_LEVEL is in { D: 70,000 - 89,999 J: 190,000 - 249,999 K: 250,000 - 299,999 } AND
CUST_MARITAL_STATUS is in { Mabsent Separ. married single widow }

Note: The carriage control characters will cause problems if you import the rules into a database table.

View Histogram

To see a histogram of the distribution of the target values in the node, select a node and click the Target Values tab at the bottom of the display.

Surrogates

Decision Trees are sensitive to missing values when applied to new data. For example, if a split in the tree (and therefore an element in the rule determining the prediction) uses the attribute Household_size, and Household_size is missing in a record to be scored, then the scoring might fail. However, if the splitting attribute is missing, the Oracle Data Mining Decision Tree algorithm provides an alternative attribute, the surrogate, to be used in the place of the missing attribute. A surrogate is another attribute that is somewhat correlated to the missing attribute; it is not always possible to find a surrogate. If both the splitting attribute and its surrogate are missing, the predicted value is determined at the parent node of the tree.

Surrogates are used during model apply. If the antecedent of a rule is missing so the rule cannot be applied, the surrogate is used in place of the rule.

To view the surrogate for a rule, select Surrogate in Split Rules at the bottom of the display. If there is a surrogate, it is displayed.

Sorting

Select Show Leaves Only. The rules for the leaf nodes (terminal nodes) are displayed in the Nodes grid. To sort rules, click the appropriate column title. For example, to sort the rules alphabetically, click Predicate; the rules are sorted in alphabetical order. To sort the rules in the reverse order, click again. Columns of numerical data, such as node ID, predicted value, confidence, cases, and support, are sorted in numerical order.

Note: You can sort rules for leaf nodes only. You cannot sort all rules.

Results

To see results, click the Results tab.

The tab contains a tree view of results. To view specific results, select the result, and click View. To delete a result, select one or more results and click Delete.

Build Settings

This tab displays the following information about how the model was built:

Oracle Data Mining 11g models can be built using Automatic Data Preparation (ADP). If the model was built using ADP, then Automatic Data Preparation has the value yes.

The following algorithm settings are displayed:

For information about Decision Tree algorithm settings, see Decision Tree Algorithm Settings.

The attributes used to build the model are displayed in the Attributes grid; the following information is displayed for each attribute:

To save the attributes information as a text file, click the Save As icon (a yellow arrow above a graphic of a page) above the Attributes grid.

Task

This tab describes the task used to build the model. It shows when the task ran, what input was used, and what output was created.