public MultiSplitDecisionTreeModelBuilder(
ISplitQualityChecker splitQualityChecker,
IBestSplitSelector bestSplitSelector,
ILeafBuilder leafBuilder,
IStatisticalSignificanceChecker statisticalSignificanceChecker = null)
: base(splitQualityChecker, bestSplitSelector, leafBuilder, statisticalSignificanceChecker)
{
}
public MultiSplitDecisionTreeModelBuilder(
ISplitQualityChecker splitQualityChecker,
IBestSplitSelector bestSplitSelector,
ILeafBuilder leafBuilder,
IStatisticalSignificanceChecker statisticalSignificanceChecker = null)
: base(splitQualityChecker, bestSplitSelector, leafBuilder, statisticalSignificanceChecker)
{
}
protected BaseDecisionTreeModelBuilder(
ISplitQualityChecker splitQualityChecker,
IBestSplitSelector bestSplitSelector,
ILeafBuilder leafBuilder,
IStatisticalSignificanceChecker statisticalSignificanceChecker = null)
{
SplitQualityChecker = splitQualityChecker;
BestSplitSelector = bestSplitSelector;
LeafBuilder = leafBuilder;
this.StatisticalSignificanceChecker = statisticalSignificanceChecker;
}
protected BaseDecisionTreeModelBuilder(
ISplitQualityChecker splitQualityChecker,
IBestSplitSelector bestSplitSelector,
ILeafBuilder leafBuilder,
IStatisticalSignificanceChecker statisticalSignificanceChecker = null)
{
SplitQualityChecker = splitQualityChecker;
BestSplitSelector = bestSplitSelector;
LeafBuilder = leafBuilder;
this.StatisticalSignificanceChecker = statisticalSignificanceChecker;
}
public BestSplitSelectorsTests()
{
ICategoricalImpurityMeasure<string> shannonEntropy = new ShannonEntropy<string>();
IBinaryDataSplitter binaryDataSplitter = new BinaryDiscreteDataSplitter();
binaryNumericBestSplitPointSelector = new ClassBreakpointsNumericSplitFinder();
binaryBestSplitSelector = new BinarySplitSelectorForCategoricalOutcome(binaryDataSplitter, binaryNumericDataSplitter, binaryNumericBestSplitPointSelector);
categoricalBinarySplitQualityChecker = new InformationGainCalculator<string>(shannonEntropy, shannonEntropy);
categoricalMultiValueSplitQualityChecker = new InformationGainCalculator<string>(shannonEntropy, shannonEntropy);
multiValueCategoricalDataSplitter = new MultiValueDiscreteDataSplitter();
multiValueBestSplitSelector = new MultiValueSplitSelectorForCategoricalOutcome(multiValueCategoricalDataSplitter, binaryNumericDataSplitter, binaryNumericBestSplitPointSelector);
dynamicProgrammingBestNumericSplitFinder = new DynamicProgrammingNumericSplitFinder();
}
public BestSplitSelectorsTests()
{
ICategoricalImpurityMeasure <string> shannonEntropy = new ShannonEntropy <string>();
IBinaryDataSplitter binaryDataSplitter = new BinaryDiscreteDataSplitter();
binaryNumericBestSplitPointSelector = new ClassBreakpointsNumericSplitFinder();
binaryBestSplitSelector = new BinarySplitSelectorForCategoricalOutcome(binaryDataSplitter, binaryNumericDataSplitter, binaryNumericBestSplitPointSelector);
categoricalBinarySplitQualityChecker = new InformationGainCalculator <string>(shannonEntropy, shannonEntropy);
categoricalMultiValueSplitQualityChecker = new InformationGainCalculator <string>(shannonEntropy, shannonEntropy);
multiValueCategoricalDataSplitter = new MultiValueDiscreteDataSplitter();
multiValueBestSplitSelector = new MultiValueSplitSelectorForCategoricalOutcome(multiValueCategoricalDataSplitter, binaryNumericDataSplitter, binaryNumericBestSplitPointSelector);
dynamicProgrammingBestNumericSplitFinder = new DynamicProgrammingNumericSplitFinder();
}
private IDecisionTreeModelBuilder BuildCustomModelBuilder(
bool binary = false,
ISplitQualityChecker splitQualityChecker = null,
IBestSplitSelector bestSplitSelector = null,
ILeafBuilder leafBuilder = null,
IStatisticalSignificanceChecker statisticalSignificanceChecker = null)
{
if (binary)
{
return(new BinaryDecisionTreeModelBuilder(
splitQualityChecker ?? new InformationGainRatioCalculator <string>(shannonEntropy, shannonEntropy as ICategoricalImpurityMeasure <string>),
bestSplitSelector as IBinaryBestSplitSelector ?? new BinarySplitSelectorForCategoricalOutcome(new BinaryDiscreteDataSplitter(), new BinaryNumericDataSplitter(), new ClassBreakpointsNumericSplitFinder()),
leafBuilder ?? new CategoricalDecisionTreeLeafBuilder(),
statisticalSignificanceChecker));
}
return(new MultiSplitDecisionTreeModelBuilder(
splitQualityChecker ?? new InformationGainRatioCalculator <string>(shannonEntropy, shannonEntropy as ICategoricalImpurityMeasure <string>),
bestSplitSelector ?? new MultiValueSplitSelectorForCategoricalOutcome(new MultiValueDiscreteDataSplitter(), new BinaryNumericDataSplitter(), new DynamicProgrammingNumericSplitFinder()),
leafBuilder ?? new CategoricalDecisionTreeLeafBuilder(),
statisticalSignificanceChecker));
}
private IDecisionTreeModelBuilder BuildCustomModelBuilder(
bool binary = false,
ISplitQualityChecker splitQualityChecker = null,
IBestSplitSelector bestSplitSelector = null,
ILeafBuilder leafBuilder = null,
IStatisticalSignificanceChecker statisticalSignificanceChecker = null)
{
if (binary)
{
return new BinaryDecisionTreeModelBuilder(
splitQualityChecker ?? new InformationGainRatioCalculator<string>(shannonEntropy, shannonEntropy as ICategoricalImpurityMeasure<string>),
bestSplitSelector as IBinaryBestSplitSelector ?? new BinarySplitSelectorForCategoricalOutcome(new BinaryDiscreteDataSplitter(), new BinaryNumericDataSplitter(), new ClassBreakpointsNumericSplitFinder()),
leafBuilder ?? new CategoricalDecisionTreeLeafBuilder(),
statisticalSignificanceChecker);
}
return new MultiSplitDecisionTreeModelBuilder(
splitQualityChecker ?? new InformationGainRatioCalculator<string>(shannonEntropy, shannonEntropy as ICategoricalImpurityMeasure<string>),
bestSplitSelector ?? new MultiValueSplitSelectorForCategoricalOutcome(new MultiValueDiscreteDataSplitter(), new BinaryNumericDataSplitter(), new DynamicProgrammingNumericSplitFinder()),
leafBuilder ?? new CategoricalDecisionTreeLeafBuilder(),
statisticalSignificanceChecker);
}
