com.spss.psapi.interaction

Interface DecisionListTaskFactory



  • public interface DecisionListTaskFactory
    Task factory for a DecisionListModelInteractor.
    • Method Detail

      • createEvaluateModelTask

        Task createEvaluateModelTask(DecisionListModel model,
                                   DecisionListTaskFactory.Partition[] partitions)
        Creates a task which will evaluate a model against a specified set of data partitions. To evaluate a model means to compute performance measures for the model and for each rule in the model. Measures are computed separately for each specified partition.

        The task must be executed on the interactor which provided this factory. If execution completes successfully the result of the task is an array of DecisionListModel containing one model for each partition. Each model in the array is a copy of the original model but with its performance measures updated.

        Parameters:
        model - the model to be evaluated
        partitions - the list of partitions against which to evaluate the model
        Returns:
        an evaluate task
      • createExtendModelTask

        Task createExtendModelTask(DecisionListModel model,
                                 DecisionListTaskFactory.Partition partition,
                                 PropertiedObject buildSettings)
        Creates a task which will extend a model against a specified data partition. To extend a model means to append new rules using the Decision List algorithm. The algorithm parameters are taken from the buildSettings object which will typically be a ModelBuilder node of type ProcessorType.BUILD_DECISION_LIST; settings not specified in this object are left as default.

        The task must be executed on the interactor which provided this factory. If execution completes successfully the result of the task is an array of DecisionListModel where each model in the array is an extension of the original model. There may be no models if the algorithm could find no additional rules satisfying the build criteria, or more than one if the algorithm found alternatives. Models are ordered in decreasing order of probability.

        Parameters:
        model - the model to be extended
        partition - the training partition
        buildSettings - specifies the algorithm parameters
        Returns:
        an extend task
      • createExtendModelTask

        Task createExtendModelTask(DecisionListModel model,
                                 DecisionListTaskFactory.Partition partition,
                                 java.util.Map buildSettings)
                                   throws InvalidPropertyException
        Creates a task which will extend a model against a specified data partition. The algorithm parameters are specified by the buildSettings which map property names to values; settings not specified in the map are left as default.

        This functions identically to createExtendModelTask(DecisionListModel, Partition, PropertiedObject) except that the properties are specified in a separate map rather than directly on the build node.

        Parameters:
        model - the model to be extended
        partition - the training partition
        buildSettings - specifies the algorithm parameters
        Returns:
        an extend task
        Throws:
        InvalidPropertyException - if the build settings are invalid
      • createDiscreteCondition

        DecisionListTaskFactory.Condition createDiscreteCondition(java.lang.String name,
                                                                java.lang.Object[] values)
        Returns a condition which selects records satisfying the predicate:
         column IN values
         
        If the list of values is empty the condition selects no records.
        Parameters:
        name - the name of the column to be tested; the column must exist in the data model and have discrete (categorical) measure
        values - a list of constant values against which the column value is to be compared; the values must be valid for the column
        Returns:
        a new selection condition

(C) Copyright IBM Corp. 1994, 2015. All Rights Reserved.