Use the decision tree algorithm to organize a collection of data belonging to several different classes or types. In the build phase, you specify a set of data vectors and provide the class of each. In the apply phase, you provide a set of previously unknown vectors and the algorithm deduces their classes from the model.
The algorithm constructs a series of simple tests or predicates to create a tree structure. To determine the class of a data vector, the algorithm takes the input data and traverses the tree from the root to the leaves performing a test at each branch.
For a categorical attribute, the test is the selection of a single attribute value out of all possible attribute values. For a numerical attribute, the test is a comparison of the attribute value with a certain threshold, precomputed by the algorithm. The outcome of the test is either assignment of the target class attribute or the selection of another attribute for further testing. The search always results in a final target assignment. The purpose of the algorithm is the construction of the most compact tree possible, while keeping the number of misclassifications at a minimum.
The decision tree algorithm enables you to specify both categorical and numerical accessors, or only categorical accessors or only numerical accessors.
You can provide both categorical and numeric predictors to this algorithm, or one or the other alone. The targets specify known classes, which are used to train the model. During the apply phase, the algorithm deduces the classes for the predictors.
The following tables show the required and optional predictors and targets for the decision tree algorithm.
Table 4. Categorical Predictors
Member Expression | Sample |
|---|---|
CategoricalPredictor.CategorialPredictor specifies member or member set to use for the predictor domain. | {[Digital], [Analog]} |
CategoricalPredictor.Sequence defines the sequence to be traversed for the predictor, generally a time dimension range. | {[Jan 1].Level.Members} |
CategoricalPredictor.External (optional) defines the scope of the predictor. | (none available) |
CategoricalPredictor.Anchor (optional) specifies additional restrictions from other dimensions. | {([2001], [Actual], [Sales], [San_Francisco])} |
Table 5. Numerical Predictors
Member Expression | Sample |
|---|---|
NumericalPredictor.NumericalPredictor specifies member or member set to use for the predictor domain. | {[Wireless]} |
NumericalPredictor.Sequence defines the sequence to be traversed for the predictor, generally a time dimension range. | {[Jan 1].Level.Members} |
NumericalPredictor.External (optional) defines the scope of the predictor. | (none available) |
NumericalPredictor.Anchor (optional) specifies additional restrictions from other dimensions. | {([2001], [Actual], [Sales], [San_Francisco])} |
Table 6. Targets
Target Expression | Example |
|---|---|
Target.Target specifies single member to use for the target domain. | {[Phones]} |
Target.Sequence defines the sequence to be traversed for the target, generally a time dimension range. | {[Jan 1].Level.Members} |
Target.External (optional) defines the scope of the target. | (none available) |
Target.Anchor (optional) specifies additional restrictions from other dimensions. | {([2001], [Actual], [Sales], [San_Francisco])} |
This information will be made available at a later release.
This information will be made available at a later release.