Package com.tibco.patterns.learn.rlink

Class ConfRatioLabel

  • All Implemented Interfaces:
    Confidence
    public class ConfRatioLabel
extends java.lang.Object
implements Confidence
    Label based confidence measure. Immutable class.

    This measure is available for model versions RFV5 and higher (ibi Patterns 5.4). It takes into account the feature scores seen and the prediction made by the model. The output confidence is in interval [0; 1]. Values below 0.5 indicate that the individual training pairs seen for this general score region disagreed with the prediction made by the model. This measure does not take into account the exact combination of feature scores.

    • Constructor Summary

      Constructors 
      Constructor Description
      ConfRatioLabel()
      Create an instance with default values.
      ConfRatioLabel​(int confidentNExamples, double confidentPercentExamples, double minCorrectPercent)
      Create an instance with specific parameter values.

    • Method Summary

      All Methods Instance Methods Concrete Methods 
      Modifier and Type Method Description
      int getConfType()
      Return the confidence type.
      double getDblArg​(int idx)
      Return a double parameter.
      int getIntArg​(int idx)
      Return an integer parameter.
      java.lang.String toString()  

      • Methods inherited from class java.lang.Object

        clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

    • Constructor Detail

      • ConfRatioLabel

        public ConfRatioLabel()
        Create an instance with default values.

      • ConfRatioLabel

        public ConfRatioLabel​(int confidentNExamples,
                      double confidentPercentExamples,
                      double minCorrectPercent)
        Create an instance with specific parameter values.
        Parameters:
        confidentNExamples - - the minimum number of supporting training examples (minus any conflicting examples) needed to reach full confidence for a feature value region. Not all examples train all regions, so this is not a count of the examples in the training set. This counts only the examples seen on the first training iteration.
        confidentPercentExamples - - the minimum percentage of supporting training examples (minus any conflicting examples) needed to reach full confidence for a feature value region. 100% represents the total number of examples seen for the same subset of feature values.
        minCorrectPercent - - percent of remaining correct examples. Larger values reduce noise in confidence values for feature scores near 0.0 or 1.0.

    • Method Detail

      • toString

        public java.lang.String toString()
        Overrides:
        toString in class java.lang.Object

      • getConfType

        public int getConfType()
        Return the confidence type.
        Specified by:
        getConfType in interface Confidence
        Returns:
        the confidence type for the Label measure.

      • getIntArg

        public int getIntArg​(int idx)
        Return an integer parameter. The first parameter is the minimum confidence count.
        Specified by:
        getIntArg in interface Confidence
        Parameters:
        idx - the integer parameter index, zero based.
        Returns:
        The minimum confidence count for index 0, -1 otherwise.

      • getDblArg

        public double getDblArg​(int idx)
        Return a double parameter. The confidence percent and minimum correct percent values.
        Specified by:
        getDblArg in interface Confidence
        Parameters:
        idx - the double parameter index, zero based.
        Returns:
        index 0: the confidence percent, index 1: the minimum correct percentage, -1.0 for all other indexes.