public void GiniClasificationImpurityCalculator_NodeImpurity()
{
var values = new double[] { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 1, 1, 2, 2, 2, 1, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, };
var unique = values.Distinct().ToArray();
var parentInterval = Interval1D.Create(0, values.Length);
var sut = new GiniClasificationImpurityCalculator();
sut.Init(unique, values, new double[0], parentInterval);
sut.UpdateIndex(50);
var actual = sut.NodeImpurity();
Assert.AreEqual(0.66666666666666674, actual, 0.000001);
}
public void GiniClasificationImpurityCalculator_LeafValue_Weighted()
{
var values = new double[] { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 1, 1, 2, 2, 2, 1, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, };
var unique = values.Distinct().ToArray();
var weights = values.Select(t => Weight(t)).ToArray();
var parentInterval = Interval1D.Create(0, values.Length);
var sut = new GiniClasificationImpurityCalculator();
sut.Init(unique, values, weights, parentInterval);
var impurity = sut.NodeImpurity();
sut.UpdateIndex(50);
var actual = sut.LeafValue();
Assert.AreEqual(2.0, actual, 0.000001);
}
public void GiniClasificationImpurityCalculator_ChildImpurities()
{
var values = new double[] { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 1, 1, 2, 2, 2, 1, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, };
var unique = values.Distinct().ToArray();
var parentInterval = Interval1D.Create(0, values.Length);
var sut = new GiniClasificationImpurityCalculator();
sut.Init(unique, values, new double[0], parentInterval);
var impurity = sut.NodeImpurity();
sut.UpdateIndex(50);
var actual = sut.ChildImpurities();
var expected = new ChildImpurities(0.0, .5);
Assert.AreEqual(expected, actual);
}
public void GiniClasificationImpurityCalculator_ImpurityImprovement()
{
var values = new double[] { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 1, 1, 2, 2, 2, 1, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, };
var unique = values.Distinct().ToArray();
var parentInterval = Interval1D.Create(0, values.Length);
var sut = new GiniClasificationImpurityCalculator();
sut.Init(unique, values, new double[0], parentInterval);
var impurity = sut.NodeImpurity();
sut.UpdateIndex(50);
var improvement1 = sut.ImpurityImprovement(impurity);
Assert.AreEqual(0.33333333333333343, improvement1, 0.000001);
sut.UpdateIndex(96);
var improvement2 = sut.ImpurityImprovement(impurity);
Assert.AreEqual(0.28047839506172845, improvement2, 0.000001);
}
public void GiniClasificationImpurityCalculator_ImpurityImprovement_Weighted()
{
var values = new double[] { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 1, 1, 2, 2, 2, 1, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, };
var unique = values.Distinct().ToArray();
var weights = values.Select(t => Weight(t)).ToArray();
var parentInterval = Interval1D.Create(0, values.Length);
var sut = new GiniClasificationImpurityCalculator();
sut.Init(unique, values, weights, parentInterval);
var impurity = sut.NodeImpurity();
sut.UpdateIndex(50);
var improvement1 = sut.ImpurityImprovement(impurity);
Assert.AreEqual(0.14015151515151511, improvement1, 0.000001);
sut.UpdateIndex(96);
var improvement2 = sut.ImpurityImprovement(impurity);
Assert.AreEqual(0.17358104858104859, improvement2, 0.000001);
}
public void RandomSplitSearcher_FindBestSplit()
{
var parser = new CsvParser(() => new StringReader(Resources.AptitudeData));
var feature = parser.EnumerateRows("AptitudeTestScore").ToF64Vector();
var targets = parser.EnumerateRows("Pass").ToF64Vector();
var interval = Interval1D.Create(0, feature.Length);
Array.Sort(feature, targets);
var impurityCalculator = new GiniClasificationImpurityCalculator();
impurityCalculator.Init(targets.Distinct().ToArray(), targets, new double[0], interval);
var impurity = impurityCalculator.NodeImpurity();
var sut = new RandomSplitSearcher(1, 42);
var actual = sut.FindBestSplit(impurityCalculator, feature, targets,
interval, impurity);
var expected = new SplitResult(15, 3.6724258636461693, 0.037941545633853213,
0.39111111111111119, 0.49586776859504134);
Assert.AreEqual(expected, actual);
}
