Model Builder

The ModelBuilder element defines the behavior of a model builder node. This is done by means of the element attributes and one or more child elements.

Format

<ModelBuilder allowNoInputs="true_false" allowNoOutputs="true_false" nullifyBlanks="true_false" 
     miningFunctions="[function1 function2 ... ]" >
     <Algorithm ... />
     <ModelingFields ... />
     <ModelGeneration ... />
     <ModelFields ... />
     <AutoModeling ... />
</ModelBuilder>

where:

allowNoInputs and allowNoOutputs must be used explicitly in a case where you want to build a model that has either no input fields or no output fields, respectively.
nullifyBlanks, if set to false, turns off the feature whereby blank values are replaced with null values (represented by $null$ ) in the data that is passed to the Model Builder component of IBM® SPSS® Modeler. The default is to replace blanks with nulls, but you might want to deactivate this feature, for example, if your algorithm needs to treat blanks differently from nulls.
miningFunctions (required) identifies the data mining function or functions that the model performs.

Table 1. Data mining functions
Function	Description
`classification`	Predicts a discrete (that is, `set`, `flag`, or `orderedSet` data type) value of an unknown target attribute from records with known target values.
`approximation`	Predicts a continuous (that is, `range` data type) value of an unknown target attribute from records with known target values.
`clustering`	Identifies groups of similar records and labels them accordingly.
`association`	Identifies related events or attributes in data.
`sequence`	Searches for sequential patterns in time-structured data.
`reduction`	Reduces the complexity of data—for example, via derived fields that summarize the contents of the original data fields.
`conceptExtraction`	Used in text mining.
`categorize`	Used in text mining.
`timeSeries`	Forecasts future values from patterns in past data.
`anomalyDetection`	Searches for unusual cases based on deviations from the norms of their cluster groups.
`attributeImportance`	Identifies the attributes that have the greatest influence on a target attribute.
`supervisedMultiTarget`	Estimates the likelihood of a (yes or no) outcome for one of a number of possibilities.

If the model performs more than one function, the function names are separated by spaces within the square brackets, as in the following example:

<ModelBuilder miningFunctions="[classification approximation]">
 ...
</ModelBuilder>

Child Elements

The child elements of the ModelBuilder element are shown in the following table.

Table 2. Child elements of model builder declaration
Child element	Description	See...
`Algorithm`	(required) Specifies the algorithm used to generate the model.	Algorithm
`ModelingFields`	Specifies the identifier to be used subsequently in the User Interface section to define the location of the input and output field controls for the model. The controls themselves are defined in the `InputFields` and `OutputFields` child elements of `ModelingFields`.	Modeling Fields
`ModelGeneration`	Specifies the identifier to be used subsequently in the User Interface section to define the location of the model name controls for the generated model.	Model Generation
`ModelFields`	Specifies the set of input and output fields used for scoring data with this model.	Model Fields
`AutoModeling`	Enables this model to be used by an ensemble modeling node, such as Auto Classifier, Auto Cluster, or Auto Numeric.	Automated Modeling