public void RegularizedLinearRegression_ArtificaialFunction()
{
// Given
var splitter = new CrossValidator <double>();
Func <IList <double>, double> scoreFunc = list => 0.3 + (0.5 * list[0]) + (-0.3 * list[1]) + (0.7 * list[2]);
var allData =
TestDataBuilder.BuildRandomAbstractNumericDataFrame(
scoreFunc,
featuresCount: 3,
min: 0,
max: 1,
rowCount: 1000);
var subject = new RegularizedLinearRegressionModelBuilder(0.5);
var regParams = new LinearRegressionParams(0.05);
// When
var accuracies = splitter.CrossValidate(
modelBuilder: subject,
modelBuilderParams: regParams,
predictor: new LinearRegressionPredictor(),
qualityMeasure: new GoodnessOfFitQualityMeasure(),
dataFrame: allData,
dependentFeatureName: "result",
percetnagOfTrainData: 0.8,
folds: 20);
// Then
Assert.IsTrue(accuracies.Select(acc => acc.Accuracy).Average() >= 0.9);
}
public void TestRegressionAndModelLeafBuilder_RandomizedData()
{
// Given
// Data generated with equation: 3 + (data["x1"] * 2) + (data["x2"] * 3) + (data["x3"] * 4)
Func <IList <double>, double> resultCalc = list => 3 + (list[0] * 2) + (list[1] * 3) + (list[2] * 4);
var testData = TestDataBuilder.BuildRandomAbstractNumericDataFrame(resultCalc, 100000, 3, min: 1, max: 100);
var leafBuilder = new RegressionAndModelDecisionTreeLeafBuilder(new GradientDescentModelBuilder(-5, 5, iterCount: 3000));
// When
var result = leafBuilder.BuildLeaf(testData, "result") as IDecisionTreeRegressionAndModelLeaf;
//Then
Assert.AreEqual(3.0, result.ModelWeights[0], 0.01);
Assert.AreEqual(2.0, result.ModelWeights[1], 0.01);
Assert.AreEqual(3.0, result.ModelWeights[2], 0.01);
Assert.AreEqual(4.0, result.ModelWeights[3], 0.01);
}