public void TestCompare()
{
var kFoldRun = new KFoldRun(10);
var experimentPerformance1 =
kFoldRun.Execute(new Classification.Experiment.Experiment(new C45(), new C45Parameter(1, true, 0.2), iris));
var experimentPerformance2 = kFoldRun.Execute(new Classification.Experiment.Experiment(new LinearPerceptron(),
new LinearPerceptronParameter(1, 0.1, 0.99, 0.2, 100), iris));
var pairedt = new Pairedt();
Assert.AreEqual(0.136, pairedt.Compare(experimentPerformance1, experimentPerformance2).GetPValue(), 0.001);
experimentPerformance1 =
kFoldRun.Execute(new Classification.Experiment.Experiment(new C45(), new C45Parameter(1, true, 0.2), tictactoe));
experimentPerformance2 =
kFoldRun.Execute(new Classification.Experiment.Experiment(new Bagging(), new BaggingParameter(1, 50), tictactoe));
Assert.AreEqual(0.00000006, pairedt.Compare(experimentPerformance1, experimentPerformance2).GetPValue(),
0.00000001);
experimentPerformance1 = kFoldRun.Execute(new Classification.Experiment.Experiment(new Lda(), new Parameter(1), dermatology));
experimentPerformance2 = kFoldRun.Execute(new Classification.Experiment.Experiment(new LinearPerceptron(),
new LinearPerceptronParameter(1, 0.1, 0.99, 0.2, 100), dermatology));
Assert.AreEqual(0.2935, pairedt.Compare(experimentPerformance1, experimentPerformance2).GetPValue(), 0.0001);
experimentPerformance1 = kFoldRun.Execute(new Classification.Experiment.Experiment(new Dummy(), new Parameter(1), nursery));
experimentPerformance2 = kFoldRun.Execute(new Classification.Experiment.Experiment(new NaiveBayes(), new Parameter(1), nursery));
Assert.AreEqual(0.0, pairedt.Compare(experimentPerformance1, experimentPerformance2).GetPValue(), 0.0000001);
experimentPerformance1 = kFoldRun.Execute(new Classification.Experiment.Experiment(new NaiveBayes(), new Parameter(1), car));
experimentPerformance2 = kFoldRun.Execute(new Classification.Experiment.Experiment(new Bagging(), new BaggingParameter(1, 50), car));
Assert.AreEqual(0.0000098, pairedt.Compare(experimentPerformance1, experimentPerformance2).GetPValue(),
0.0000001);
experimentPerformance1 =
kFoldRun.Execute(new Classification.Experiment.Experiment(new Knn(), new KnnParameter(1, 3, new EuclidianDistance()), bupa));
experimentPerformance2 = kFoldRun.Execute(new Classification.Experiment.Experiment(new Lda(), new Parameter(1), bupa));
Assert.AreEqual(0.1020, pairedt.Compare(experimentPerformance1, experimentPerformance2).GetPValue(), 0.0001);
}
public void TestExecute()
{
var kFoldRun = new KFoldRun(10);
var experimentPerformance =
kFoldRun.Execute(new Classification.Experiment.Experiment(new C45(), new C45Parameter(1, true, 0.2), iris));
Assert.AreEqual(6.00, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance =
kFoldRun.Execute(new Classification.Experiment.Experiment(new C45(), new C45Parameter(1, true, 0.2), tictactoe));
Assert.AreEqual(16.39, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance =
kFoldRun.Execute(new Classification.Experiment.Experiment(new Knn(), new KnnParameter(1, 3, new EuclidianDistance()), bupa));
Assert.AreEqual(37.44, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new Knn(),
new KnnParameter(1, 3, new EuclidianDistance()), dermatology));
Assert.AreEqual(9.59, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new Lda(), new Parameter(1), bupa));
Assert.AreEqual(31.83, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new Lda(), new Parameter(1), dermatology));
Assert.AreEqual(2.18, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new LinearPerceptron(),
new LinearPerceptronParameter(1, 0.1, 0.99, 0.2, 100), iris));
Assert.AreEqual(2.67, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new LinearPerceptron(),
new LinearPerceptronParameter(1, 0.1, 0.99, 0.2, 100), dermatology));
Assert.AreEqual(4.89, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new NaiveBayes(), new Parameter(1), car));
Assert.AreEqual(14.64, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new NaiveBayes(), new Parameter(1), nursery));
Assert.AreEqual(9.71, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance =
kFoldRun.Execute(new Classification.Experiment.Experiment(new Bagging(), new BaggingParameter(1, 50), tictactoe));
Assert.AreEqual(3.03, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new Bagging(), new BaggingParameter(1, 50), car));
Assert.AreEqual(6.25, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new Dummy(), new Parameter(1), nursery));
Assert.AreEqual(67.17, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
experimentPerformance = kFoldRun.Execute(new Classification.Experiment.Experiment(new Dummy(), new Parameter(1), iris));
Assert.AreEqual(80.00, 100 * experimentPerformance.MeanPerformance().GetErrorRate(), 0.01);
}
