public static double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, IClassificationProblemData problemData, IEnumerable <int> rows, bool applyLinearScaling)
{
IEnumerable <double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows);
IEnumerable <double> targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
OnlineCalculatorError errorState;
double mse;
if (applyLinearScaling)
{
var mseCalculator = new OnlineMeanSquaredErrorCalculator();
CalculateWithScaling(targetValues, estimatedValues, lowerEstimationLimit, upperEstimationLimit, mseCalculator, problemData.Dataset.Rows);
errorState = mseCalculator.ErrorState;
mse = mseCalculator.MeanSquaredError;
}
else
{
IEnumerable <double> boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
mse = OnlineMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
}
if (errorState != OnlineCalculatorError.None)
{
return(Double.NaN);
}
return(mse);
}
public static double Calculate(ISymbolicTimeSeriesPrognosisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, ITimeSeriesPrognosisProblemData problemData, IEnumerable<int> rows, IntRange evaluationPartition, int horizon, bool applyLinearScaling) {
var horizions = rows.Select(r => Math.Min(horizon, evaluationPartition.End - r));
IEnumerable<double> targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows.Zip(horizions, Enumerable.Range).SelectMany(r => r));
IEnumerable<double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows, horizions).SelectMany(x => x);
OnlineCalculatorError errorState;
double mse;
if (applyLinearScaling && horizon == 1) { //perform normal evaluation and afterwards scale the solution and calculate the fitness value
var mseCalculator = new OnlineMeanSquaredErrorCalculator();
CalculateWithScaling(targetValues, estimatedValues, lowerEstimationLimit, upperEstimationLimit, mseCalculator, problemData.Dataset.Rows * horizon);
errorState = mseCalculator.ErrorState;
mse = mseCalculator.MeanSquaredError;
} else if (applyLinearScaling) { //first create model to perform linear scaling and afterwards calculate fitness for the scaled model
var model = new SymbolicTimeSeriesPrognosisModel((ISymbolicExpressionTree)solution.Clone(), interpreter, lowerEstimationLimit, upperEstimationLimit);
model.Scale(problemData);
var scaledSolution = model.SymbolicExpressionTree;
estimatedValues = interpreter.GetSymbolicExpressionTreeValues(scaledSolution, problemData.Dataset, rows, horizions).SelectMany(x => x);
var boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
mse = OnlineMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
} else {
var boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
mse = OnlineMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
}
if (errorState != OnlineCalculatorError.None) return Double.NaN;
else return mse;
}
// calculate variable relevance based on removal of variables
// 1) to remove a variable we set it's length scale to infinity (no relation of the variable value to the target)
// 2) calculate MSE of the original target values (y) to the updated targes y' (after variable removal)
// 3) relevance is larger if MSE(y,y') is large
// 4) scale impacts so that the most important variable has impact = 1
private double[] CalculateRelevance(double[] y, double[] u, List <double>[] xs, double[] l)
{
int nRows = xs.First().Count;
var changedL = new double[l.Length];
var relevance = new double[l.Length];
for (int i = 0; i < l.Length; i++)
{
Array.Copy(l, changedL, changedL.Length);
changedL[i] = double.MaxValue;
var changedK = CalculateCovariance(xs, changedL);
var yChanged = new double[u.Length];
alglib.ablas.rmatrixmv(nRows, nRows, changedK, 0, 0, 0, u, 0, ref yChanged, 0);
OnlineCalculatorError error;
var mse = OnlineMeanSquaredErrorCalculator.Calculate(y, yChanged, out error);
if (error != OnlineCalculatorError.None)
{
mse = double.MaxValue;
}
relevance[i] = mse;
}
// scale so that max relevance is 1.0
var maxRel = relevance.Max();
for (int i = 0; i < relevance.Length; i++)
{
relevance[i] /= maxRel;
}
return(relevance);
}
// calculate variable relevance based on removal of variables
// 1) to remove a variable we set it's coefficient to zero
// 2) calculate MSE of the original target values (y) to the updated targes y' (after variable removal)
// 3) relevance is larger if MSE(y,y') is large
// 4) scale impacts so that the most important variable has impact = 1
private double[] CalculateRelevance(double[] y, List <double>[] xs, double[] l)
{
var changedL = new double[l.Length];
var relevance = new double[l.Length];
for (int i = 0; i < l.Length; i++)
{
Array.Copy(l, changedL, changedL.Length);
changedL[i] = 0.0;
var yChanged = EvaluteLinearModel(xs, changedL);
OnlineCalculatorError error;
var mse = OnlineMeanSquaredErrorCalculator.Calculate(y, yChanged, out error);
if (error != OnlineCalculatorError.None)
{
mse = double.MaxValue;
}
relevance[i] = mse;
}
// scale so that max relevance is 1.0
var maxRel = relevance.Max();
for (int i = 0; i < relevance.Length; i++)
{
relevance[i] /= maxRel;
}
return(relevance);
}
public static double[] Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, IClassificationProblemData problemData, IEnumerable <int> rows)
{
IEnumerable <double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows);
IEnumerable <double> originalValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
IEnumerable <double> boundedEstimationValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
OnlineCalculatorError errorState;
double mse = OnlineMeanSquaredErrorCalculator.Calculate(originalValues, boundedEstimationValues, out errorState);
if (errorState != OnlineCalculatorError.None)
{
mse = double.NaN;
}
return(new double[2] {
mse, solution.Length
});
}
public static double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, IClassificationProblemData problemData, IEnumerable<int> rows, bool applyLinearScaling) {
IEnumerable<double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows);
IEnumerable<double> targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
OnlineCalculatorError errorState;
double mse;
if (applyLinearScaling) {
var mseCalculator = new OnlineMeanSquaredErrorCalculator();
CalculateWithScaling(targetValues, estimatedValues, lowerEstimationLimit, upperEstimationLimit, mseCalculator, problemData.Dataset.Rows);
errorState = mseCalculator.ErrorState;
mse = mseCalculator.MeanSquaredError;
} else {
IEnumerable<double> boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
mse = OnlineMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
}
if (errorState != OnlineCalculatorError.None) return Double.NaN;
return mse;
}
private static double CalculateCrossValidationPartitions(Tuple <svm_problem, svm_problem>[] partitions, svm_parameter parameters)
{
double avgTestMse = 0;
var calc = new OnlineMeanSquaredErrorCalculator();
foreach (Tuple <svm_problem, svm_problem> tuple in partitions)
{
var trainingSvmProblem = tuple.Item1;
var testSvmProblem = tuple.Item2;
var model = svm.svm_train(trainingSvmProblem, parameters);
calc.Reset();
for (int i = 0; i < testSvmProblem.l; ++i)
{
calc.Add(testSvmProblem.y[i], svm.svm_predict(model, testSvmProblem.x[i]));
}
double mse = calc.ErrorState == OnlineCalculatorError.None ? calc.MeanSquaredError : double.NaN;
avgTestMse += mse;
}
avgTestMse /= partitions.Length;
return(avgTestMse);
}
public static double Calculate(ISymbolicTimeSeriesPrognosisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, ITimeSeriesPrognosisProblemData problemData, IEnumerable <int> rows, IntRange evaluationPartition, int horizon, bool applyLinearScaling)
{
var horizions = rows.Select(r => Math.Min(horizon, evaluationPartition.End - r));
IEnumerable <double> targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows.Zip(horizions, Enumerable.Range).SelectMany(r => r));
IEnumerable <double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows, horizions).SelectMany(x => x);
OnlineCalculatorError errorState;
double mse;
if (applyLinearScaling && horizon == 1) //perform normal evaluation and afterwards scale the solution and calculate the fitness value
{
var mseCalculator = new OnlineMeanSquaredErrorCalculator();
CalculateWithScaling(targetValues, estimatedValues, lowerEstimationLimit, upperEstimationLimit, mseCalculator, problemData.Dataset.Rows * horizon);
errorState = mseCalculator.ErrorState;
mse = mseCalculator.MeanSquaredError;
}
else if (applyLinearScaling) //first create model to perform linear scaling and afterwards calculate fitness for the scaled model
{
var model = new SymbolicTimeSeriesPrognosisModel(problemData.TargetVariable, (ISymbolicExpressionTree)solution.Clone(), interpreter, lowerEstimationLimit, upperEstimationLimit);
model.Scale(problemData);
var scaledSolution = model.SymbolicExpressionTree;
estimatedValues = interpreter.GetSymbolicExpressionTreeValues(scaledSolution, problemData.Dataset, rows, horizions).SelectMany(x => x);
var boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
mse = OnlineMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
}
else
{
var boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
mse = OnlineMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
}
if (errorState != OnlineCalculatorError.None)
{
return(Double.NaN);
}
else
{
return(mse);
}
}
private static void CrossValidate(IRegressionProblemData problemData, Tuple <IEnumerable <int>, IEnumerable <int> >[] partitions, int nTrees, double r, double m, int seed, out double avgTestMse)
{
avgTestMse = 0;
var ds = problemData.Dataset;
var targetVariable = GetTargetVariableName(problemData);
foreach (var tuple in partitions)
{
double rmsError, avgRelError, outOfBagAvgRelError, outOfBagRmsError;
var trainingRandomForestPartition = tuple.Item1;
var testRandomForestPartition = tuple.Item2;
var model = RandomForestModel.CreateRegressionModel(problemData, trainingRandomForestPartition, nTrees, r, m, seed, out rmsError, out avgRelError, out outOfBagRmsError, out outOfBagAvgRelError);
var estimatedValues = model.GetEstimatedValues(ds, testRandomForestPartition);
var targetValues = ds.GetDoubleValues(targetVariable, testRandomForestPartition);
OnlineCalculatorError calculatorError;
double mse = OnlineMeanSquaredErrorCalculator.Calculate(estimatedValues, targetValues, out calculatorError);
if (calculatorError != OnlineCalculatorError.None)
{
mse = double.NaN;
}
avgTestMse += mse;
}
avgTestMse /= partitions.Length;
}
protected override void Run()
{
IRegressionProblemData problemData = Problem.ProblemData;
IEnumerable <string> selectedInputVariables = problemData.AllowedInputVariables;
int nSv;
ISupportVectorMachineModel model;
Run(problemData, selectedInputVariables, SvmType.Value, KernelType.Value, Cost.Value, Nu.Value, Gamma.Value, Epsilon.Value, Degree.Value, out model, out nSv);
if (CreateSolution)
{
var solution = new SupportVectorRegressionSolution((SupportVectorMachineModel)model, (IRegressionProblemData)problemData.Clone());
Results.Add(new Result("Support vector regression solution", "The support vector regression solution.", solution));
}
Results.Add(new Result("Number of support vectors", "The number of support vectors of the SVR solution.", new IntValue(nSv)));
{
// calculate regression model metrics
var ds = problemData.Dataset;
var trainRows = problemData.TrainingIndices;
var testRows = problemData.TestIndices;
var yTrain = ds.GetDoubleValues(problemData.TargetVariable, trainRows);
var yTest = ds.GetDoubleValues(problemData.TargetVariable, testRows);
var yPredTrain = model.GetEstimatedValues(ds, trainRows).ToArray();
var yPredTest = model.GetEstimatedValues(ds, testRows).ToArray();
OnlineCalculatorError error;
var trainMse = OnlineMeanSquaredErrorCalculator.Calculate(yPredTrain, yTrain, out error);
if (error != OnlineCalculatorError.None)
{
trainMse = double.MaxValue;
}
var testMse = OnlineMeanSquaredErrorCalculator.Calculate(yPredTest, yTest, out error);
if (error != OnlineCalculatorError.None)
{
testMse = double.MaxValue;
}
Results.Add(new Result("Mean squared error (training)", "The mean of squared errors of the SVR solution on the training partition.", new DoubleValue(trainMse)));
Results.Add(new Result("Mean squared error (test)", "The mean of squared errors of the SVR solution on the test partition.", new DoubleValue(testMse)));
var trainMae = OnlineMeanAbsoluteErrorCalculator.Calculate(yPredTrain, yTrain, out error);
if (error != OnlineCalculatorError.None)
{
trainMae = double.MaxValue;
}
var testMae = OnlineMeanAbsoluteErrorCalculator.Calculate(yPredTest, yTest, out error);
if (error != OnlineCalculatorError.None)
{
testMae = double.MaxValue;
}
Results.Add(new Result("Mean absolute error (training)", "The mean of absolute errors of the SVR solution on the training partition.", new DoubleValue(trainMae)));
Results.Add(new Result("Mean absolute error (test)", "The mean of absolute errors of the SVR solution on the test partition.", new DoubleValue(testMae)));
var trainRelErr = OnlineMeanAbsolutePercentageErrorCalculator.Calculate(yPredTrain, yTrain, out error);
if (error != OnlineCalculatorError.None)
{
trainRelErr = double.MaxValue;
}
var testRelErr = OnlineMeanAbsolutePercentageErrorCalculator.Calculate(yPredTest, yTest, out error);
if (error != OnlineCalculatorError.None)
{
testRelErr = double.MaxValue;
}
Results.Add(new Result("Average relative error (training)", "The mean of relative errors of the SVR solution on the training partition.", new DoubleValue(trainRelErr)));
Results.Add(new Result("Average relative error (test)", "The mean of relative errors of the SVR solution on the test partition.", new DoubleValue(testRelErr)));
}
}
