public void Test_GetNumericColumnVector_ByIndex()
{
// Given
var expectedNumericColumnVector = Vector <double> .Build.Dense(new double[] { 1, 3, 5 });
// When
var actualNumericColumnVector = _subject.GetNumericColumnVector(1);
// Then
Assert.IsTrue(expectedNumericColumnVector.Equals(actualNumericColumnVector));
}
private static List <NumericFeatureData> OrderColumn(IDataFrame baseData, string dependentFeatureName, string numericFeatureToProcess)
{
return(baseData.GetNumericColumnVector(numericFeatureToProcess)
.Select(
(featureVal, rowIndex) =>
new NumericFeatureData(featureVal, baseData[rowIndex, dependentFeatureName].FeatureValue))
.OrderBy(elem => elem.FeatureVal)
.ToList());
}
public Tuple <ISplittingResult, double> FindBestSplitPoint(
IDataFrame baseData,
string dependentFeatureName,
string numericFeatureToProcess,
ICategoricalSplitQualityChecker splitQualityChecker,
IBinaryNumericDataSplitter binaryNumericDataSplitter,
double initialEntropy)
{
ISplittingResult bestSplit = null;
double bestSplitQuality = double.NegativeInfinity;
var totalRowsCount = baseData.RowCount;
var sortedRowData =
baseData.GetNumericColumnVector(numericFeatureToProcess)
.AsParallel()
.Select((val, rowIdx) =>
new
{
RowIdx = rowIdx,
FeatureValue = val,
DependentFeatureValue = baseData[rowIdx, dependentFeatureName].FeatureValue
})
.OrderBy(elem => elem.FeatureValue).ThenBy(elem => elem.RowIdx)
.ToList();
var previousClass = sortedRowData[0].DependentFeatureValue;
var previousFeatureVal = sortedRowData[0].FeatureValue;
foreach (var rowData in sortedRowData)
{
var currentClass = rowData.DependentFeatureValue;
var currentFeatureVal = rowData.FeatureValue;
if (!currentClass.Equals(previousClass) && !currentFeatureVal.Equals(previousFeatureVal))
{
var halfWay = (previousFeatureVal + currentFeatureVal) / 2.0;
var splitParams = new BinarySplittingParams(numericFeatureToProcess, halfWay, dependentFeatureName);
var splitResult = binaryNumericDataSplitter.SplitData(baseData, splitParams);
var quality = splitQualityChecker.CalculateSplitQuality(
initialEntropy,
totalRowsCount,
splitResult,
dependentFeatureName);
if (quality >= bestSplitQuality)
{
bestSplitQuality = quality;
bestSplit = new BinarySplittingResult(true, numericFeatureToProcess, splitResult, halfWay);
}
previousClass = currentClass;
}
previousFeatureVal = currentFeatureVal;
}
return(new Tuple <ISplittingResult, double>(bestSplit, bestSplitQuality));
}
public IPredictionModel BuildModel(IDataFrame dataFrame, string dependentFeatureName, IModelBuilderParams additionalParams)
{
if (!(additionalParams is ILinearRegressionParams))
{
throw new ArgumentException("Invalid parameters passed to Regularized Linear Regression model builder!");
}
var linearRegressionParams = additionalParams as ILinearRegressionParams;
var matrixX = dataFrame.GetSubsetByColumns(dataFrame.ColumnNames.Except(new[] { dependentFeatureName }).ToList()).GetAsMatrixWithIntercept();
var vectorY = dataFrame.GetNumericColumnVector(dependentFeatureName);
return BuildModel(matrixX, vectorY, linearRegressionParams);
}
public IPredictionModel BuildModel(IDataFrame dataFrame, string dependentFeatureName, IModelBuilderParams additionalParams)
{
if (!(additionalParams is ILinearRegressionParams))
{
throw new ArgumentException("Invalid parameters passed to Gradient Desccent model builder!");
}
var linearRegressionParams = additionalParams as ILinearRegressionParams;
var matrixX = dataFrame.GetSubsetByColumns(dataFrame.ColumnNames.Except(new[] { dependentFeatureName }).ToList()).GetAsMatrixWithIntercept();
var vectorY = dataFrame.GetNumericColumnVector(dependentFeatureName);
return(BuildModel(matrixX, vectorY, linearRegressionParams));
}
public Tuple<ISplittingResult, double> FindBestSplitPoint(
IDataFrame baseData,
string dependentFeatureName,
string numericFeatureToProcess,
ICategoricalSplitQualityChecker splitQualityChecker,
IBinaryNumericDataSplitter binaryNumericDataSplitter,
double initialEntropy)
{
ISplittingResult bestSplit = null;
double bestSplitQuality = double.NegativeInfinity;
var totalRowsCount = baseData.RowCount;
var sortedRowData =
baseData.GetNumericColumnVector(numericFeatureToProcess)
.AsParallel()
.Select((val, rowIdx) =>
new
{
RowIdx = rowIdx,
FeatureValue = val,
DependentFeatureValue = baseData[rowIdx, dependentFeatureName].FeatureValue
})
.OrderBy(elem => elem.FeatureValue).ThenBy(elem => elem.RowIdx)
.ToList();
var previousClass = sortedRowData[0].DependentFeatureValue;
var previousFeatureVal = sortedRowData[0].FeatureValue;
foreach (var rowData in sortedRowData)
{
var currentClass = rowData.DependentFeatureValue;
var currentFeatureVal = rowData.FeatureValue;
if (!currentClass.Equals(previousClass) && !currentFeatureVal.Equals(previousFeatureVal))
{
var halfWay = (previousFeatureVal + currentFeatureVal) / 2.0;
var splitParams = new BinarySplittingParams(numericFeatureToProcess, halfWay, dependentFeatureName);
var splitResult = binaryNumericDataSplitter.SplitData(baseData, splitParams);
var quality = splitQualityChecker.CalculateSplitQuality(
initialEntropy,
totalRowsCount,
splitResult,
dependentFeatureName);
if (quality >= bestSplitQuality)
{
bestSplitQuality = quality;
bestSplit = new BinarySplittingResult(true, numericFeatureToProcess, splitResult, halfWay);
}
previousClass = currentClass;
}
previousFeatureVal = currentFeatureVal;
}
return new Tuple<ISplittingResult, double>(bestSplit, bestSplitQuality);
}
public IDecisionTreeLeaf BuildLeaf(IDataFrame finalData, string dependentFeatureName)
{
var vectorY = finalData.GetNumericColumnVector(dependentFeatureName);
var featureNames = finalData.ColumnNames.Except(new[] { dependentFeatureName }).ToList();
var subset = finalData.GetSubsetByColumns(featureNames);
var matrixX = finalData.GetSubsetByColumns(featureNames).GetAsMatrixWithIntercept();
Vector <double> fittedWeights = null;
try
{
fittedWeights = MultipleRegression.DirectMethod(matrixX, vectorY);
}
catch (Exception)
{
fittedWeights = regressionModelBuilder.BuildModel(matrixX, vectorY, regressionParams).Weights;
}
return(new RegressionAndModelLeaf(dependentFeatureName, fittedWeights, vectorY.Mean()));
}
public IDecisionTreeLeaf BuildLeaf(IDataFrame finalData, string dependentFeatureName)
{
var vectorY = finalData.GetNumericColumnVector(dependentFeatureName);
var featureNames = finalData.ColumnNames.Except(new[] { dependentFeatureName }).ToList();
var subset = finalData.GetSubsetByColumns(featureNames);
var matrixX = finalData.GetSubsetByColumns(featureNames).GetAsMatrixWithIntercept();
Vector<double> fittedWeights = null;
try
{
fittedWeights = MultipleRegression.DirectMethod(matrixX, vectorY);
}
catch (Exception)
{
fittedWeights = regressionModelBuilder.BuildModel(matrixX, vectorY, regressionParams).Weights;
}
return new RegressionAndModelLeaf(dependentFeatureName, fittedWeights, vectorY.Mean());
}
public double GetInitialEntropy(IDataFrame baseData, string dependentFeatureName)
{
return(baseData.GetNumericColumnVector(dependentFeatureName).Variance());
}
public ISplittingResult SelectBestSplit(
IDataFrame baseData,
string dependentFeatureName,
INumericalSplitQualityChecker splitQualityChecker,
IAlredyUsedAttributesInfo alreadyUsedAttributesInfo)
{
var bestSplitQuality = double.NegativeInfinity;
var initialEntropy = splitQualityChecker.GetInitialEntropy(baseData, dependentFeatureName);
Tuple <string, double> bestSplit = null;
/*
* if (baseData.RowCount <= baseData.ColumnsCount)
* {
* return null;
* }
*/
var featureColumns = baseData.ColumnNames.Except(new[] { dependentFeatureName });
foreach (var feature in featureColumns)
{
var dataOrderedByFeature =
baseData.GetNumericColumnVector(feature)
.Select((rowVal, idx) => new Tuple <double, double, int>(rowVal, (double)baseData[idx, dependentFeatureName].FeatureValue, idx))
.OrderBy(tpl => tpl.Item1)
.ToList();
var dependentFeatureValuesOrdered = dataOrderedByFeature.Select(elem => elem.Item2).ToList();
var previousFeatureValue = dataOrderedByFeature.First().Item1;
for (int i = 0; i < (dataOrderedByFeature.Count - 1); i++)
{
var dataPoint = dataOrderedByFeature[i];
var currentFeatureValue = dataPoint.Item1;
if (currentFeatureValue != previousFeatureValue)
{
var splitPoint = (currentFeatureValue + previousFeatureValue) / 2.0;
if (!alreadyUsedAttributesInfo.WasAttributeAlreadyUsedWithValue(feature, splitPoint))
{
var dependentValsBelow = dependentFeatureValuesOrdered.Take(i).ToList();
var dependentValsAbove = dependentFeatureValuesOrdered.Skip(i).ToList();
var splitQuality = splitQualityChecker.CalculateSplitQuality(
initialEntropy,
baseData.RowCount,
new[] { dependentValsBelow, dependentValsAbove });
if (splitQuality > bestSplitQuality)
{
bestSplitQuality = splitQuality;
bestSplit = new Tuple <string, double>(feature, splitPoint);
}
}
}
previousFeatureValue = currentFeatureValue;
}
}
if (bestSplit == null)
{
return(null);
}
var splittedData = binaryNumericDataSplitter.SplitData(
baseData,
new BinarySplittingParams(bestSplit.Item1, bestSplit.Item2, dependentFeatureName));
return(new BinarySplittingResult(true, bestSplit.Item1, splittedData, bestSplit.Item2));
}
public double GetInitialEntropy(IDataFrame baseData, string dependentFeatureName)
{
return baseData.GetNumericColumnVector(dependentFeatureName).Variance();
}
private static List<NumericFeatureData> OrderColumn(IDataFrame baseData, string dependentFeatureName, string numericFeatureToProcess)
{
return baseData.GetNumericColumnVector(numericFeatureToProcess)
.Select(
(featureVal, rowIndex) =>
new NumericFeatureData(featureVal, baseData[rowIndex, dependentFeatureName].FeatureValue))
.OrderBy(elem => elem.FeatureVal)
.ToList();
}
public ISplittingResult SelectBestSplit(
IDataFrame baseData,
string dependentFeatureName,
INumericalSplitQualityChecker splitQualityChecker,
IAlredyUsedAttributesInfo alreadyUsedAttributesInfo)
{
var bestSplitQuality = double.NegativeInfinity;
var initialEntropy = splitQualityChecker.GetInitialEntropy(baseData, dependentFeatureName);
Tuple<string, double> bestSplit = null;
/*
if (baseData.RowCount <= baseData.ColumnsCount)
{
return null;
}
*/
var featureColumns = baseData.ColumnNames.Except(new[] { dependentFeatureName });
foreach (var feature in featureColumns)
{
var dataOrderedByFeature =
baseData.GetNumericColumnVector(feature)
.Select((rowVal, idx) => new Tuple<double, double, int>(rowVal, (double)baseData[idx, dependentFeatureName].FeatureValue, idx))
.OrderBy(tpl => tpl.Item1)
.ToList();
var dependentFeatureValuesOrdered = dataOrderedByFeature.Select(elem => elem.Item2).ToList();
var previousFeatureValue = dataOrderedByFeature.First().Item1;
for (int i = 0; i < (dataOrderedByFeature.Count -1); i++)
{
var dataPoint = dataOrderedByFeature[i];
var currentFeatureValue = dataPoint.Item1;
if (currentFeatureValue != previousFeatureValue)
{
var splitPoint = (currentFeatureValue + previousFeatureValue) / 2.0;
if (!alreadyUsedAttributesInfo.WasAttributeAlreadyUsedWithValue(feature, splitPoint))
{
var dependentValsBelow = dependentFeatureValuesOrdered.Take(i).ToList();
var dependentValsAbove = dependentFeatureValuesOrdered.Skip(i).ToList();
var splitQuality = splitQualityChecker.CalculateSplitQuality(
initialEntropy,
baseData.RowCount,
new[] { dependentValsBelow, dependentValsAbove });
if (splitQuality > bestSplitQuality)
{
bestSplitQuality = splitQuality;
bestSplit = new Tuple<string, double>(feature, splitPoint);
}
}
}
previousFeatureValue = currentFeatureValue;
}
}
if (bestSplit == null)
{
return null;
}
var splittedData = binaryNumericDataSplitter.SplitData(
baseData,
new BinarySplittingParams(bestSplit.Item1, bestSplit.Item2, dependentFeatureName));
return new BinarySplittingResult(true, bestSplit.Item1, splittedData, bestSplit.Item2);
}
