public static IKrigingModel selectModel(KrigingModelType value)
{
IKrigingModel model;
switch (value)
{
case KrigingModelType.circular:
model = new ModelCircular();
break;
case KrigingModelType.exponential:
model = new ModelExponential();
break;
case KrigingModelType.gaussian:
model = new ModelGaussian();
break;
case KrigingModelType.linear:
model = new ModelLineal();
break;
case KrigingModelType.spherical:
model = new ModelSpherical();
break;
default:
model = new ModelExponential();
break;
}
return(model);
}
public IKrigingModel setFunction(KrigingModelType model, double sill, double range, double nugget)
{
_krigingModel = selectModel(model);
_krigingModel.Nugget = nugget;
_krigingModel.Sill = sill;
_krigingModel.Range = range;
onChangedKriginParameter(EventArgs.Empty);
return(_krigingModel);
}
public IKrigingModel getFunctionDefault(KrigingModelType model)
{
return(getFunction(model, _spatial.getSemivarianceStatistic().Mean,
_spatial.getLag().lagSize *_spatial.getLag().numLags / 2, 0));
}
public IKrigingModel getFunction(KrigingModelType model, double sill, double range, double nugget)
{
_msill = new Parameter(sill);
_mrange = new Parameter(range);
_mnugget = new Parameter(nugget);
_distance = new Parameter();
// Parameter[] regressionParameters=null;
// Parameter[] observedParameters = null; ;
_observedParameters = new[] { _distance };
// regressionParameters = new Parameter[] { mrange, msill,mnugget };
defineParametersCalculation();
switch (model)
{
case KrigingModelType.circular:
_krigingModel = new ModelCircular();
_regressionFunction = () =>
_distance > 0 && _distance <= Math.Abs(_mrange)
? _mnugget +
(((2 * Math.Abs(_msill)) / Math.PI) *
(((_distance / Math.Abs(_mrange)) *
Math.Sqrt(1 - ((_distance * _distance) / (Math.Abs(_mrange) * Math.Abs(_mrange))))) +
Math.Asin(_distance / Math.Abs(_mrange))))
: _distance > Math.Abs(_mrange) ? _mnugget + Math.Abs(_msill) : 0
;
break;
case KrigingModelType.exponential:
_krigingModel = new ModelExponential();
_regressionFunction = () => _distance > 0 ?
Math.Abs(_mnugget) + (Math.Abs(_msill) * (1 - Math.Exp(-3 * Math.Abs(_distance) / Math.Abs(_mrange))))
: 0;
break;
case KrigingModelType.gaussian:
_krigingModel = new ModelGaussian();
_regressionFunction = () => _distance > 0 ?
Math.Abs(_mnugget) +
(Math.Abs(_msill) *
(1 -
Math.Exp(-3 * (Math.Abs(_distance) / Math.Abs(_mrange)) * (Math.Abs(_distance) / Math.Abs(_mrange)))))
: 0;
break;
case KrigingModelType.spherical:
_krigingModel = new ModelSpherical();
_regressionFunction = () =>
_distance > 0 && _distance <= Math.Abs(_mrange) ?
Math.Abs(_mnugget) + (Math.Abs(_msill) * ((1.5 * (_distance / Math.Abs(_mrange))) -
(0.5 * (_distance / Math.Abs(_mrange)) *
(_distance / Math.Abs(_mrange)) *
(_distance / Math.Abs(_mrange))))) :
_distance > Math.Abs(_mrange) ?
Math.Abs(_mnugget) + Math.Abs(_msill) :
0;
break;
case KrigingModelType.linear:
_krigingModel = new ModelLineal();
_regressionFunction = () =>
_distance < _mrange ? _mnugget + ((_msill / _mrange) * Math.Abs(_distance)): _mnugget + _msill;
break;
}
if (_regressionParameters.Length == 0)
{
_krigingModel.Range = range;
_krigingModel.Sill = sill;
if (nugget < 0)
{
_krigingModel.Nugget = 0;
}
else
{
_krigingModel.Nugget = nugget;
}
onChangedKriginParameter(EventArgs.Empty);
return(_krigingModel);
}
if (_regressionParameters != null)
{
try {
var levenbergMarquardt = new LevenbergMarquardt(_regressionFunction, _regressionParameters,
_observedParameters, _z);
for (int i = 0; i < 50; i++)
{
levenbergMarquardt.Iterate();
}
defineParametersCalculationValues(sill, range, nugget);
} catch {
_krigingModel.Range = range;
_krigingModel.Sill = sill;
if (nugget < 0)
{
_krigingModel.Nugget = 0;
}
else
{
_krigingModel.Nugget = nugget;
}
onChangedKriginParameter(EventArgs.Empty);
return(_krigingModel);
}
}
onChangedKriginParameter(EventArgs.Empty);
return(_krigingModel);
}
