com.imsl.datamining.decisionTree

Class DecisionTreeInfoGain

java.lang.Object

com.imsl.datamining.PredictiveModel

com.imsl.datamining.decisionTree.DecisionTree

com.imsl.datamining.decisionTree.DecisionTreeInfoGain

All Implemented Interfaces:

Serializable, Cloneable

Direct Known Subclasses:

ALACART, C45

public abstract class DecisionTreeInfoGain
extends DecisionTree
implements Serializable, Cloneable

Abstract class that extends DecisionTree for classes that use an information gain criteria.

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	DecisionTreeInfoGain.GainCriteria Specifies which information gain criteria to use in determining the best split at each node.

Nested classes/interfaces inherited from class com.imsl.datamining.decisionTree.DecisionTree

DecisionTree.MaxTreeSizeExceededException, DecisionTree.PruningFailedToConvergeException, DecisionTree.PureNodeException

Nested classes/interfaces inherited from class com.imsl.datamining.PredictiveModel

PredictiveModel.PredictiveModelException, PredictiveModel.StateChangeException, PredictiveModel.SumOfProbabilitiesNotOneException, PredictiveModel.VariableType

Constructor Summary

Constructors

Constructor and Description
DecisionTreeInfoGain(double[][] xy, int responseColumnIndex, PredictiveModel.VariableType[] varType) Constructs a DecisionTree object for a single response variable and multiple predictor variables.

Method Summary

Methods

Modifier and Type	Method and Description
protected double	information(int[] x, int[] y, double[] classCounts, double[] weights, boolean xInfo) Returns the expected information of a variable y over a partition determined by the variable x.
protected abstract int	selectSplitVariable(double[][] xy, double[] classCounts, double[] parentFreq, double[] splitValue, int[] splitPartition) Abstract method for selecting the next split variable and split definition for the node.
void	setGainCriteria(DecisionTreeInfoGain.GainCriteria gainCriteria) Specifies which criteria to use in gain calculations in order to determine the best split at each node.
void	setUseRatio(boolean ratio) Sets the flag to use or not use the gain ratio instead of the gain to determine the best split.
boolean	useGainRatio() Returns whether or not the gain ratio is to be used instead of the gain to determine the best split.

Methods inherited from class com.imsl.datamining.decisionTree.DecisionTree

fitModel, getCostComplexityValues, getDecisionTree, getFittedMeanSquaredError, getMaxDepth, getMaxNodes, getMeanSquaredPredictionError, getMinObsPerChildNode, getMinObsPerNode, getNumberOfComplexityValues, getNumberOfSets, isAutoPruningFlag, predict, predict, predict, printDecisionTree, printDecisionTree, pruneTree, setAutoPruningFlag, setConfiguration, setCostComplexityValues, setMaxDepth, setMaxNodes, setMinCostComplexityValue, setMinObsPerChildNode, setMinObsPerNode

Methods inherited from class com.imsl.datamining.PredictiveModel

getClassCounts, getCostMatrix, getMaxNumberOfCategories, getNumberOfClasses, getNumberOfColumns, getNumberOfMissing, getNumberOfPredictors, getNumberOfRows, getNumberOfUniquePredictorValues, getPredictorIndexes, getPredictorTypes, getPrintLevel, getPriorProbabilities, getResponseColumnIndex, getResponseVariableAverage, getResponseVariableMostFrequentClass, getResponseVariableType, getTotalWeight, getVariableType, getWeights, getXY, isMustFitModelFlag, isUserFixedNClasses, setClassCounts, setCostMatrix, setMaxNumberOfCategories, setNumberOfClasses, setPredictorIndex, setPredictorTypes, setPrintLevel, setPriorProbabilities, setWeights

Methods inherited from class java.lang.Object

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

Constructor Detail

DecisionTreeInfoGain

public DecisionTreeInfoGain(double[][] xy,
                    int responseColumnIndex,
                    PredictiveModel.VariableType[] varType)

Constructs a DecisionTree object for a single response variable and multiple predictor variables.

Parameters:

xy - a double matrix with rows containing the observations on the predictor variables and one response variable.

responseColumnIndex - an int specifying the column index of the response variable.

varType - a PredictiveModel.VariableType array containing the type of each variable.

Method Detail

information

protected double information(int[] x,
                 int[] y,
                 double[] classCounts,
                 double[] weights,
                 boolean xInfo)

Returns the expected information of a variable y over a partition determined by the variable x.

Given a data subset containing both variables and , let

be a partition of determined by the values in . Then the expected information is

where is either the Shannon entropy or the Gini index, according to DecisionTreeInfoGain.GainCriteria. Note: if is constant, the return value is the Shannon Entropy (or Gini index) of Y.

Parameters:

x - an int array of length xy.length containing values of a predictor or an indicator vector defining the partition of the observations.

y - int array of length xy.length containing the values of the response variable.

classCounts - a double array containing the counts for each class of the response variable, when it is categorical.

weights - a double array used to indicate which subset of the observations belong in the current node.

xInfo - a boolean indicating that we are getting information about x using a simple frequency estimate.

Value	Method
true	simple frequency estimate
false	prior probabilities

Returns:

a double indicating the information uncertainty.

selectSplitVariable

protected abstract int selectSplitVariable(double[][] xy,
                      double[] classCounts,
                      double[] parentFreq,
                      double[] splitValue,
                      int[] splitPartition)

Abstract method for selecting the next split variable and split definition for the node.

Specified by:

selectSplitVariable in class DecisionTree

Parameters:

xy - a double matrix containing the data.

classCounts - a double array containing the counts for each class of the response variable, when it is categorical.

parentFreq - a double array used to indicate which subset of the observations belong in the current node.

splitValue - a double array representing the resulting split point if the selected variable is quantitative.

splitPartition - an int array indicating the resulting split partition if the selected variable is categorical.

Returns:

an int specifying the column index of the split variable in xy.

setGainCriteria

public void setGainCriteria(DecisionTreeInfoGain.GainCriteria gainCriteria)

Specifies which criteria to use in gain calculations in order to determine the best split at each node.

Parameters:

gainCriteria - a DecisionTreeInfoGain.GainCriteria specifying which criteria to use in gain calculations in order to determine the best split at each node.

Default: gainCriteria = DecisionTreeInfoGain.GainCriteria.SHANNON_ENTROPY

setUseRatio

public void setUseRatio(boolean ratio)

Sets the flag to use or not use the gain ratio instead of the gain to determine the best split.

Parameters:

ratio - a boolean indicating if the gain ratio is to be used.

true results in the gain ratio being used and false indicates the gain is to be used.

Default: useRatio=false

useGainRatio

public boolean useGainRatio()

Returns whether or not the gain ratio is to be used instead of the gain to determine the best split.

Returns:

a boolean indicating if the gain ratio is to be used.

true results in the gain ratio being used and false indicates the gain is to be used.