public void RegressionBackwardEliminationModelSelectingEnsembleLearner_CreateMetaFeatures_Then_Learn()
{
var learners = new IIndexedLearner <double>[]
{
new RegressionDecisionTreeLearner(2),
new RegressionDecisionTreeLearner(5),
new RegressionDecisionTreeLearner(7),
new RegressionDecisionTreeLearner(9),
new RegressionDecisionTreeLearner(11),
new RegressionDecisionTreeLearner(21),
new RegressionDecisionTreeLearner(23),
new RegressionDecisionTreeLearner(1),
new RegressionDecisionTreeLearner(14),
new RegressionDecisionTreeLearner(17),
new RegressionDecisionTreeLearner(19),
new RegressionDecisionTreeLearner(33)
};
var sut = new RegressionBackwardEliminationModelSelectingEnsembleLearner(learners, 5);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var metaObservations = sut.LearnMetaFeatures(observations, targets);
var model = sut.SelectModels(observations, metaObservations, targets);
var predictions = model.Predict(observations);
var metric = new MeanSquaredErrorRegressionMetric();
var actual = metric.Error(targets, predictions);
Assert.AreEqual(0.010316259438112841, actual, 0.0001);
}
public void RegressionStackingEnsembleLearner_CreateMetaFeatures_Then_Learn()
{
var learners = new IIndexedLearner <double>[]
{
new RegressionDecisionTreeLearner(2),
new RegressionDecisionTreeLearner(5),
new RegressionDecisionTreeLearner(7),
new RegressionDecisionTreeLearner(9)
};
var sut = new RegressionStackingEnsembleLearner(learners, new RegressionDecisionTreeLearner(9),
new RandomCrossValidation <double>(5, 23), false);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var metaObservations = sut.LearnMetaFeatures(observations, targets);
var model = sut.LearnStackingModel(observations, metaObservations, targets);
var predictions = model.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.06951934687172627, actual, 0.0001);
}
public void GBMDecisionTreeLearner_Learn()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var inSample = targets.Select(t => true).ToArray();
var orderedElements = new int[observations.ColumnCount][];
var rows = observations.RowCount;
for (int i = 0; i < observations.ColumnCount; i++)
{
var feature = observations.Column(i);
var indices = Enumerable.Range(0, rows).ToArray();
feature.SortWith(indices);
orderedElements[i] = indices;
}
var sut = new GBMDecisionTreeLearner(10);
var tree = sut.Learn(observations, targets, targets, targets, orderedElements, inSample);
var predictions = tree.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.0046122425037232661, actual);
}
public void RegressionStackingEnsembleLearner_Learn_Indexed()
{
var learners = new IIndexedLearner <double>[]
{
new RegressionDecisionTreeLearner(2),
new RegressionDecisionTreeLearner(5),
new RegressionDecisionTreeLearner(7),
new RegressionDecisionTreeLearner(9)
};
var sut = new RegressionStackingEnsembleLearner(learners, new RegressionDecisionTreeLearner(9),
new RandomCrossValidation <double>(5, 23), false);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var indices = Enumerable.Range(0, 25).ToArray();
var model = sut.Learn(observations, targets, indices);
var predictions = model.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.133930222950635, actual, 0.0001);
}
public void RegressionRandomModelSelectingEnsembleLearner_Learn_Indexed()
{
var learners = new IIndexedLearner <double>[]
{
new RegressionDecisionTreeLearner(2),
new RegressionDecisionTreeLearner(5),
new RegressionDecisionTreeLearner(7),
new RegressionDecisionTreeLearner(9),
new RegressionDecisionTreeLearner(11),
new RegressionDecisionTreeLearner(21),
new RegressionDecisionTreeLearner(23),
new RegressionDecisionTreeLearner(1),
new RegressionDecisionTreeLearner(14),
new RegressionDecisionTreeLearner(17),
new RegressionDecisionTreeLearner(19),
new RegressionDecisionTreeLearner(33)
};
var sut = new RegressionRandomModelSelectingEnsembleLearner(learners, 5);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var indices = Enumerable.Range(0, 25).ToArray();
var model = sut.Learn(observations, targets, indices);
var predictions = model.Predict(observations);
var metric = new MeanSquaredErrorRegressionMetric();
var actual = metric.Error(targets, predictions);
Assert.AreEqual(0.13601421174394385, actual, 0.0001);
}
public void RegressionForwardSearchModelSelectingEnsembleLearner_Learn_Without_Replacement()
{
var learners = new IIndexedLearner <double>[]
{
new RegressionDecisionTreeLearner(2),
new RegressionDecisionTreeLearner(5),
new RegressionDecisionTreeLearner(7),
new RegressionDecisionTreeLearner(9),
new RegressionDecisionTreeLearner(11),
new RegressionDecisionTreeLearner(21),
new RegressionDecisionTreeLearner(23),
new RegressionDecisionTreeLearner(1),
new RegressionDecisionTreeLearner(14),
new RegressionDecisionTreeLearner(17),
new RegressionDecisionTreeLearner(19),
new RegressionDecisionTreeLearner(33)
};
var metric = new MeanSquaredErrorRegressionMetric();
var sut = new RegressionForwardSearchModelSelectingEnsembleLearner(learners, 5,
new RandomCrossValidation <double>(5, 42),
new MeanRegressionEnsembleStrategy(), metric, 1, false);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var model = sut.Learn(observations, targets);
var predictions = model.Predict(observations);
var actual = metric.Error(targets, predictions);
Assert.AreEqual(0.010316259438112841, actual, 0.0001);
}
public void RegressionRandomModelSelectingEnsembleLearner_Learn()
{
var learners = new IIndexedLearner <double>[]
{
new RegressionDecisionTreeLearner(2),
new RegressionDecisionTreeLearner(5),
new RegressionDecisionTreeLearner(7),
new RegressionDecisionTreeLearner(9),
new RegressionDecisionTreeLearner(11),
new RegressionDecisionTreeLearner(21),
new RegressionDecisionTreeLearner(23),
new RegressionDecisionTreeLearner(1),
new RegressionDecisionTreeLearner(14),
new RegressionDecisionTreeLearner(17),
new RegressionDecisionTreeLearner(19),
new RegressionDecisionTreeLearner(33)
};
var sut = new RegressionRandomModelSelectingEnsembleLearner(learners, 5);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var model = sut.Learn(observations, targets);
var predictions = model.Predict(observations);
var metric = new MeanSquaredErrorRegressionMetric();
var actual = metric.Error(targets, predictions);
Assert.AreEqual(0.017238327841614508, actual, 0.0001);
}
public void RegressionRandomModelSelectingEnsembleLearner_Learn_Start_With_3_Models()
{
var learners = new IIndexedLearner <double>[]
{
new RegressionDecisionTreeLearner(2),
new RegressionDecisionTreeLearner(5),
new RegressionDecisionTreeLearner(7),
new RegressionDecisionTreeLearner(9),
new RegressionDecisionTreeLearner(11),
new RegressionDecisionTreeLearner(21),
new RegressionDecisionTreeLearner(23),
new RegressionDecisionTreeLearner(1),
new RegressionDecisionTreeLearner(14),
new RegressionDecisionTreeLearner(17),
new RegressionDecisionTreeLearner(19),
new RegressionDecisionTreeLearner(33)
};
var metric = new MeanSquaredErrorRegressionMetric();
var sut = new RegressionRandomModelSelectingEnsembleLearner(learners, 5,
new RandomCrossValidation <double>(5, 42), new MeanRegressionEnsembleStrategy(), metric, 3, false);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var model = sut.Learn(observations, targets);
var predictions = model.Predict(observations);
var actual = metric.Error(targets, predictions);
Assert.AreEqual(0.0090143589987671056, actual, 0.0001);
}
public void GBMTree_AddRawFeatureImportances()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var inSample = targets.Select(t => true).ToArray();
var orderedElements = new int[observations.ColumnCount][];
var rows = observations.RowCount;
for (int i = 0; i < observations.ColumnCount; i++)
{
var feature = observations.Column(i);
var indices = Enumerable.Range(0, rows).ToArray();
feature.SortWith(indices);
orderedElements[i] = indices;
}
var sut = new GBMDecisionTreeLearner(10);
var tree = sut.Learn(observations, targets, targets, targets, orderedElements, inSample);
var actual = new double[observations.ColumnCount];
tree.AddRawVariableImportances(actual);
var expected = new double[] { 0.0, 105017.48701572006 };
Assert.AreEqual(expected.Length, actual.Length);
Assert.AreEqual(expected[0], actual[0], 0.01);
Assert.AreEqual(expected[1], actual[1], 0.01);
}
public void TimeSeriesCrossValidation_Validate_InitialTrainingSize_Is_Larger_Than_Obsevations_Length()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new TimeSeriesCrossValidation <double>(initialTrainingSize: 300);
var timeSeriesPredictions = sut.Validate(new RegressionDecisionTreeLearner(), observations, targets);
}
public void TimeSeriesCrossValidation_Validate_Observations_And_Targets_Length_Does_Not_Match()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
targets = targets.Take(100).ToArray();
var sut = new TimeSeriesCrossValidation <double>(initialTrainingSize: 5);
var timeSeriesPredictions = sut.Validate(new RegressionDecisionTreeLearner(), observations, targets);
}
double CrossValidate(int folds)
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new CrossValidation <double>(new RandomIndexSampler <double>(42), folds);
var predictions = sut.CrossValidate(new RegressionDecisionTreeLearner(), observations, targets);
var metric = new MeanSquaredErrorRegressionMetric();
return(metric.Error(targets, predictions));
}
double CrossValidate(int folds)
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new NoShuffleCrossValidation <double>(folds);
var learner = new RegressionDecisionTreeLearner();
var predictions = sut.CrossValidate(learner, observations, targets);
var metric = new MeanSquaredErrorRegressionMetric();
return(metric.Error(targets, predictions));
}
public void RegressionDecisionTreeModel_Save()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var learner = new RegressionDecisionTreeLearner(100, 4, 2, 0.1, 42);
var sut = learner.Learn(observations, targets);
var writer = new StringWriter();
sut.Save(() => writer);
Assert.AreEqual(m_regressionDecisionTreeModelString, writer.ToString());
}
public void RegressionGradientBoostLearner_LearnWithEarlyStopping_ToFewIterations()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var splitter = new RandomTrainingTestIndexSplitter <double>(0.6, 1234);
var split = splitter.SplitSet(observations, targets);
var sut = new RegressionSquareLossGradientBoostLearner(5, 0.1, 3, 1, 1e-6, 1.0, 0, false);
var evaluator = new MeanSquaredErrorRegressionMetric();
var model = sut.LearnWithEarlyStopping(split.TrainingSet.Observations, split.TrainingSet.Targets,
split.TestSet.Observations, split.TestSet.Targets, evaluator, 5);
}
public void RegressionDecisionTreeModel_Load()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var reader = new StringReader(m_regressionDecisionTreeModelString);
var sut = RegressionDecisionTreeModel.Load(() => reader);
var predictions = sut.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var error = evaluator.Error(targets, predictions);
Assert.AreEqual(0.032120286249559482, error, 0.0000001);
}
public void RegressionDecisionTreeModel_Predict_Multiple()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var learner = new RegressionDecisionTreeLearner(100, 4, 2, 0.1, 42);
var sut = learner.Learn(observations, targets);
var predictions = sut.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var error = evaluator.Error(targets, predictions);
Assert.AreEqual(0.032120286249559482, error, 0.0000001);
}
public void RegressionQuantileLossGradientBoostLearner_Learn()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new RegressionQuantileLossGradientBoostLearner(50, 0.1, 3, 1, 1e-6, 1.0, 0, 0.9, false);
var model = sut.Learn(observations, targets);
var predictions = model.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.18540395091912656, actual);
}
public void RegressionAbsoluteLossGradientBoostLearner_Stochastic_Learn()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new RegressionAbsoluteLossGradientBoostLearner(50, 0.1, 3, 1, 1e-6, .5, 0, false);
var model = sut.Learn(observations, targets);
var predictions = model.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.033412842952357739, actual, 0.0001);
}
public void RegressionAbsoluteLossGradientBoostLearner_FeaturesPrSplit_Learn()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new RegressionAbsoluteLossGradientBoostLearner(50, 0.1, 3, 1, 1e-6, 1.0, 1, false);
var model = sut.Learn(observations, targets);
var predictions = model.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var actual = evaluator.Error(targets, predictions);
Assert.AreEqual(0.0861480348494789, actual);
}
public void TimeSeriesCrossValidation_Validate_MaxTrainingSetSize_And_RetrainInterval()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new TimeSeriesCrossValidation <double>(initialTrainingSize: 5, maxTrainingSetSize: 30, retrainInterval: 5);
var timeSeriesPredictions = sut.Validate(new RegressionDecisionTreeLearner(), observations, targets);
var timeSeriesTargets = sut.GetValidationTargets(targets);
var metric = new MeanSquaredErrorRegressionMetric();
var error = metric.Error(timeSeriesTargets, timeSeriesPredictions);
Assert.AreEqual(0.13010151998135897, error, 0.00001);
}
public void TimeSeriesCrossValidation_Validate_MaxTrainingSetSize()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new TimeSeriesCrossValidation <double>(initialTrainingSize: 5, maxTrainingSetSize: 10);
var timeSeriesPredictions = sut.Validate(new RegressionDecisionTreeLearner(), observations, targets);
var timeSeriesTargets = sut.GetValidationTargets(targets);
var metric = new MeanSquaredErrorRegressionMetric();
var error = metric.Error(timeSeriesTargets, timeSeriesPredictions);
Assert.AreEqual(0.27296549371962692, error, 0.00001);
}
private double RegressionDecisionTreeLearner_Learn_Weighted(int treeDepth, double weight)
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new RegressionDecisionTreeLearner(treeDepth, 4, 2, 0.1, 42);
var weights = targets.Select(v => Weight(v, weight)).ToArray();
var model = sut.Learn(observations, targets, weights);
var predictions = model.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var error = evaluator.Error(targets, predictions);
return(error);
}
public void TimeSeriesCrossValidation_Validate_RetrainInterval()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new TimeSeriesCrossValidation <double>(initialTrainingSize: 5, retrainInterval: 5);
var learner = new RegressionDecisionTreeLearner();
var timeSeriesPredictions = sut.Validate(learner, observations, targets);
var timeSeriesTargets = sut.GetValidationTargets(targets);
var metric = new MeanSquaredErrorRegressionMetric();
var error = metric.Error(timeSeriesTargets, timeSeriesPredictions);
Assert.AreEqual(0.096346937132994928, error, 0.00001);
}
double CrossValidate_Provide_Indices(int folds)
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new RandomCrossValidation <double>(folds, 42);
var rowsToCrossvalidate = targets.Length / 2;
var indices = Enumerable.Range(0, rowsToCrossvalidate).ToArray();
var predictions = new double[rowsToCrossvalidate];
sut.CrossValidate(new RegressionDecisionTreeLearner(), observations, targets, indices, predictions);
var metric = new MeanSquaredErrorRegressionMetric();
return(metric.Error(targets.Take(rowsToCrossvalidate).ToArray(), predictions));
}
private static double RegressionDecisionTreeLearner_Learn(int treeDepth)
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var sut = new RegressionDecisionTreeLearner(treeDepth, 4, 2, 0.1, 42);
var model = sut.Learn(observations, targets);
var predictions = model.Predict(observations);
var evaluator = new MeanSquaredErrorRegressionMetric();
var error = evaluator.Error(targets, predictions);
return(error);
}
public void RegressionDecisionTreeModel_Predict_Multiple_Indexed()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var learner = new RegressionDecisionTreeLearner(100, 4, 2, 0.1, 42);
var sut = learner.Learn(observations, targets);
var indices = new int[] { 0, 3, 4, 5, 6, 7, 8, 9, 20, 21 };
var predictions = sut.Predict(observations, indices);
var indexedTargets = targets.GetIndices(indices);
var evaluator = new MeanSquaredErrorRegressionMetric();
var error = evaluator.Error(indexedTargets, predictions);
Assert.AreEqual(0.023821615502626264, error, 0.0000001);
}
public void NoShuffleLearningCurvesCalculator_Calculate()
{
var sut = new NoShuffleLearningCurvesCalculator <double>(new MeanSquaredErrorRegressionMetric(),
new double[] { 0.2, 0.8 }, 0.8);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var actual = sut.Calculate(new RegressionDecisionTreeLearner(),
observations, targets);
var expected = new List <LearningCurvePoint>()
{
new LearningCurvePoint(32, 0, 0.12874833873980004),
new LearningCurvePoint(128, 0.0, 0.067720786718774989)
};
CollectionAssert.AreEqual(expected, actual);
}
public void RegressionDecisionTreeModel_GetVariableImportance()
{
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var featureNameToIndex = new Dictionary <string, int> {
{ "F1", 0 }, { "F2", 1 }
};
var learner = new RegressionDecisionTreeLearner(100, 4, 2, 0.1, 42);
var sut = learner.Learn(observations, targets);
var actual = sut.GetVariableImportance(featureNameToIndex);
var expected = new Dictionary <string, double> {
{ "F2", 100.0 }, { "F1", 0.0 }
};
CollectionAssert.AreEqual(expected, actual);
}
public void RandomLearningCurvesCalculator_Calculate()
{
var sut = new RandomShuffleLearningCurvesCalculator <double>(new MeanSquaredErrorRegressionMetric(),
new double[] { 0.2, 0.8 }, 0.8, 42, 5);
var(observations, targets) = DataSetUtilities.LoadDecisionTreeDataSet();
var actual = sut.Calculate(new RegressionDecisionTreeLearner(),
observations, targets);
var expected = new List <LearningCurvePoint>()
{
new LearningCurvePoint(32, 0, 0.141565953928265),
new LearningCurvePoint(128, 0.0, 0.068970597423950036)
};
CollectionAssert.AreEqual(expected, actual);
}