/** Method: Constructor */
internal MontecarloConv(int maxIt)
{
rg = new RndGenerator();
convData = new List <double>();
hist = new Histogram(100);
de = new KernelDensity(1, 100, 100);
this.maxIt = 2000; //TODO: eliminate parameter
}
public double DiversityDisabilities()
{
List <double> listeCount = new List <double>();
foreach (Disabilities item in Enum.GetValues(typeof(Disabilities)))
{
listeCount.Add(this.employees.Where(x => x.disabilities == item).Count());
}
return(Math.Round((1 - KernelDensity.GaussianKernel(1 / Statistics.StandardDeviation(listeCount))) * 100));
}
public double DiversityCivilState()
{
List <double> listeCount = new List <double>();
foreach (CivilState item in Enum.GetValues(typeof(CivilState)))
{
listeCount.Add(this.employees.Where(x => x.civil_State == item).Count());
}
return(Math.Round((KernelDensity.GaussianKernel(1 / Statistics.StandardDeviation(listeCount))) * 100));
}
public void KDETestGaussianKernelBandwidth2()
{
var estimate = KernelDensity.EstimateGaussian(-3.5d, 2.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.046875864115900, estimate, 10);
estimate = KernelDensity.EstimateGaussian(0.0d, 2.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.186580447512078, estimate, 10);
estimate = KernelDensity.EstimateGaussian(2.0d, 2.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.123339405007761, estimate, 10);
}
public void KDETestGaussianKernelBandwidth0p5()
{
var estimate = KernelDensity.EstimateGaussian(-3.5d, 0.5d, _testData);
AssertHelpers.AlmostEqualRelative(5.311490430807364e-007, estimate, 10);
estimate = KernelDensity.EstimateGaussian(0.0d, 0.5d, _testData);
AssertHelpers.AlmostEqualRelative(0.369994803886827, estimate, 10);
estimate = KernelDensity.EstimateGaussian(2.0d, 0.5d, _testData);
AssertHelpers.AlmostEqualRelative(0.032447347007482, estimate, 10);
}
internal MontecarloGenericConv(int maxIt, int normMin)
{
distributions = new List <Distribution>();
de = new KernelDensity(1, 100, 100);
nd = new NormalDistrib();
rg = new RndGenerator();
this.maxIt = maxIt;
this.normMin = normMin;
this.mean = -1;
this.stDev = -1;
}
public void KDETestUniformKernelBandwidth1()
{
Assert.AreEqual(0.5d, KernelDensity.UniformKernel(0));
var estimate = KernelDensity.EstimateUniform(-3.5d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0, estimate, 10);
estimate = KernelDensity.EstimateUniform(0.0d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.35, estimate, 10);
estimate = KernelDensity.EstimateUniform(2.0d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.1, estimate, 10);
}
public void KDETestTriangularKernelBandwidth1()
{
Assert.AreEqual(1.0d, KernelDensity.TriangularKernel(0));
var estimate = KernelDensity.EstimateTriangular(-3.5d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0, estimate, 10);
estimate = KernelDensity.EstimateTriangular(0.0d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.347688490533868, estimate, 10);
estimate = KernelDensity.EstimateTriangular(2.0d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.004216757636608, estimate, 10);
}
public void KDETestEpanechnikovKernelBandwidth1()
{
Assert.AreEqual(0.75d, KernelDensity.EpanechnikovKernel(0));
var estimate = KernelDensity.EstimateEpanechnikov(-3.5d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0, estimate, 10);
estimate = KernelDensity.EstimateEpanechnikov(0.0d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.353803214812608, estimate, 10);
estimate = KernelDensity.EstimateEpanechnikov(2.0d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.006248168996717, estimate, 10);
}
public double DiversityAge()
{
List <double> listeCount = new List <double>();
foreach (Age item in Enum.GetValues(typeof(Age)))
{
listeCount.Add(this.employees.Where(x => x.age == item).Count());
}
var y = Statistics.StandardDeviation(listeCount);
return(Math.Round((KernelDensity.GaussianKernel(1 / y)) * 100));
}
public void KDETestGaussianKernelBandwidth1()
{
//Density of standard normal distribution at 0
AssertHelpers.AlmostEqualRelative(0.398942280401433, KernelDensity.GaussianKernel(0), 10);
var estimate = KernelDensity.EstimateGaussian(-3.5d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.004115405028907, estimate, 10);
estimate = KernelDensity.EstimateGaussian(0.0d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.310485907659139, estimate, 10);
estimate = KernelDensity.EstimateGaussian(2.0d, 1.0d, _testData);
AssertHelpers.AlmostEqualRelative(0.099698581377801, estimate, 10);
}
public void KDETestCustomKernelBandwidth1()
{
double Kernel(double x) => 0.5d * Math.Exp(-Math.Abs(x));
Assert.AreEqual(0.5d, Kernel(0));
var estimate = KernelDensity.Estimate(-3.5d, 1.0d, _testData, Kernel);
AssertHelpers.AlmostEqualRelative(0.018396636706009, estimate, 10);
estimate = KernelDensity.Estimate(0.0d, 1.0d, _testData, Kernel);
AssertHelpers.AlmostEqualRelative(0.272675897096678, estimate, 10);
estimate = KernelDensity.Estimate(2.0d, 1.0d, _testData, Kernel);
AssertHelpers.AlmostEqualRelative(0.092580285110347, estimate, 10);
}
/** Method: Load data for calculation
* data - data from time series
* n - number (real) of convolutions */
void IConvolution.LoadData(List <double> data, double n)
{
if (data.Count == 0)
{
throw new Exception("Error. No data");
}
double res = n - (int)n;
this.n = n;
convData.Clear();
hist.Clear();
de = new KernelDensity(1, 100, 100);
int [] indexes = new int[(int)n + 1];
int indexRes = -1;
double sum;
for (int i = 0; i < maxIt / 2; i++)
{
sum = 0;
for (int j = 1; j <= n; j++)
{
indexes[j - 1] = rg.NextInt(0, data.Count - 1);
sum += data[indexes[j - 1]];
}
if (res > 0)
{
indexRes = rg.NextInt(0, data.Count - 1);
sum = sum + data[indexRes] * res;
}
convData.Add(sum);
sum = 0;
for (int j = 1; j <= n; j++)
{
sum += data[data.Count - 1 - indexes[j - 1]];
}
if (res > 0)
{
sum = sum + data[data.Count - 1 - indexRes] * res;
}
convData.Add(sum);
}
hist.LoadData(convData);
de.LoadHist(hist);
de.SetMaxInt();
}
private void LoadMontConvolution()
{
List <double> convData = new List <double>();
Histogram convHist = new Histogram();
double sum;
for (int i = 0; i < maxIt; i++)
{
sum = 0;
foreach (Distribution dist in distributions)
{
dist.RawData = dist.Histogram.GetRawData();
}
foreach (Distribution dist in distributions)
{
int nLeadTimesInt = (int)dist.NLeadTimes;
double nLeadTimesRes = dist.NLeadTimes - nLeadTimesInt;
int[] indexes = new int[(int)dist.NLeadTimes + 1];
int indexRes = -1;
for (int j = 1; j <= nLeadTimesInt; j++)
{
indexes[j - 1] = rg.NextInt(0, dist.RawData.Count - 1);
sum += dist.RawData[indexes[j - 1]];
}
if (nLeadTimesRes > 0)
{
indexRes = rg.NextInt(0, dist.RawData.Count - 1);
sum = sum + dist.RawData[indexRes] * nLeadTimesRes;
}
convData.Add(sum * dist.Factor);
}
}
convHist.LoadData(convData);
de = new KernelDensity(1, 100, 100);
de.LoadHist(convHist);
de.SetMaxInt();
}
