id3
Class TDDTOptions

java.lang.Object
  |
  +--id3.TDDTOptions

public class TDDTOptions

extends java.lang.Object

This class stores information about option settings for Top-Down Decision Tree inducers.

Field Summary

boolean	adjustThresholds TRUE indicates threshold should be adjusted to equal instance values, FALSE otherwise.
boolean	contMDLAdjust TRUE if the Minimum Description Length Adjustment for continuous attributes should be applied to mutual info, FALSE otherwise.
boolean	debug TRUE if debugging options are used.
boolean	emptyNodeParentDist TRUE indicates an empty node should have the parent's distribution, FALSE otherwise.
byte	evaluationMetric The metric used to evaluate this decision tree.
double	evidenceFactor Evidence correction factor.
byte	leafDistType Type of distribution to build at leaves.
double	lowerBoundMinSplitWeight The lower bound for the minimum weight of instances in a node.
int	maxLevel The maximum level of growth.
double	MEstimateFactor M-estimate factor for laplace.
double	minSplitWeightPercent The percent (p) used to calculate the min weight of instances in a node (m).
boolean	nominalLBoundOnly TRUE indicates lowerBoundMinSplitWeight, upperBoundMinSplitWeight, and minSplitWeightPercent are not used for setting minimum instances in a node for nominal attributes, FALSE indicates they will be used.
boolean	parentTieBreaking TRUE indicates a node should inherit the parent's tie-breaking class, FALSE otherwise.
boolean	pruningBranchReplacement TRUE indicates pruning should allow replacing a node with its largest subtree, FALSE otherwise.
double	pruningFactor Factor of how much pruning should be done.
byte	pruningMethod Pruning method to be used.
double	smoothFactor Exponential factor for smoothing.
int	smoothInst Number of thresholds on either side to use for smoothing; 0 for no smoothing.
byte	splitScoreCriterion The criterion used for scoring.
boolean	unknownEdges TRUE indicates there will be an edge with "unknown" from every node.
double	upperBoundMinSplitWeight The upper bound for the minimum weight of instances in a node.

Constructor Summary

TDDTOptions()
Constructor.

Methods inherited from class java.lang.Object

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

Field Detail

maxLevel

public int maxLevel

The maximum level of growth.

lowerBoundMinSplitWeight

public double lowerBoundMinSplitWeight

The lower bound for the minimum weight of instances in a node.

upperBoundMinSplitWeight

public double upperBoundMinSplitWeight

The upper bound for the minimum weight of instances in a node.

minSplitWeightPercent

public double minSplitWeightPercent

The percent (p) used to calculate the min weight of instances in a node (m).
m = p * num instances / num categories

nominalLBoundOnly

public boolean nominalLBoundOnly

TRUE indicates lowerBoundMinSplitWeight, upperBoundMinSplitWeight, and minSplitWeightPercent are not used for setting minimum instances in a node for nominal attributes, FALSE indicates they will be used.

debug

public boolean debug

TRUE if debugging options are used.

unknownEdges

public boolean unknownEdges

TRUE indicates there will be an edge with "unknown" from every node.

splitScoreCriterion

public byte splitScoreCriterion

The criterion used for scoring.

emptyNodeParentDist

public boolean emptyNodeParentDist

TRUE indicates an empty node should have the parent's distribution, FALSE otherwise.

parentTieBreaking

public boolean parentTieBreaking

TRUE indicates a node should inherit the parent's tie-breaking class, FALSE otherwise.

pruningMethod

public byte pruningMethod

Pruning method to be used. If the value is not NONE and pruning_factor is 0, then a node will be made a leaf when its (potential) children do not improve the error count.

pruningBranchReplacement

public boolean pruningBranchReplacement

TRUE indicates pruning should allow replacing a node with its largest subtree, FALSE otherwise.

adjustThresholds

public boolean adjustThresholds

TRUE indicates threshold should be adjusted to equal instance values, FALSE otherwise.

pruningFactor

public double pruningFactor

Factor of how much pruning should be done. High values indicate more pruning.

contMDLAdjust

public boolean contMDLAdjust

TRUE if the Minimum Description Length Adjustment for continuous attributes should be applied to mutual info, FALSE otherwise.

smoothInst

public int smoothInst

Number of thresholds on either side to use for smoothing; 0 for no smoothing.

smoothFactor

public double smoothFactor

Exponential factor for smoothing.

leafDistType

public byte leafDistType

Type of distribution to build at leaves.

MEstimateFactor

public double MEstimateFactor

M-estimate factor for laplace.

evidenceFactor

public double evidenceFactor

Evidence correction factor.

evaluationMetric

public byte evaluationMetric

The metric used to evaluate this decision tree.

Constructor Detail

TDDTOptions

public TDDTOptions()

Constructor.