public bool Import(string pathFile)
{
InitTraining();
XmlDocument input = new XmlDocument();
try
{
input.Load(pathFile);
XmlNode root = input.FirstChild;
//Get number of input, hidden, output nodes
this._regularDifferencingLevel = Int32.Parse(root.SelectSingleNode("descendant::RegularDifferencing").InnerText);
this._seasonDifferencingLevel = Int32.Parse(root.SelectSingleNode("descendant::SeasonDifferencing").InnerText);
this._seasonPartern = Int32.Parse(root.SelectSingleNode("descendant::SeasonPartern").InnerText);
this._pRegular = Int32.Parse(root.SelectSingleNode("descendant::ARRegular").InnerText);
this._qRegular = Int32.Parse(root.SelectSingleNode("descendant::MARegular").InnerText);
this._pSeason = Int32.Parse(root.SelectSingleNode("descendant::ARSeason").InnerText);
this._qSeason = Int32.Parse(root.SelectSingleNode("descendant::MASeason").InnerText);
string arimaCoef = root.SelectSingleNode("descendant::ARIMACoef").InnerText;
string[] listArimaCoef = arimaCoef.Split('|');
_listArimaCoef = new List <double>();
foreach (string coef in listArimaCoef)
{
double temp = Double.Parse(coef);
_listArimaCoef.Add(temp);
}
Statistic.ComputeDifference(ref _processARIMASeries, ref _startIndex, _regularDifferencingLevel, _seasonDifferencingLevel, _seasonPartern);
//ComputTrainingError(_processARIMASeries, _startIndex, _regularDifferencingLevel, _seasonDifferencingLevel, _pRegular, _qRegular, _seasonPartern, _pSeason, _qSeason, _listArimaCoef, out _errorARIMASeries);
}
catch
{
}
return(true);
}
public void ManualTraining(int pRegular, int regularDifferencing, int qRegular, int pSeason, int seassonDifferencing, int qSeason, int seasonPartern)
{
InitTraining();
this._regularDifferencingLevel = regularDifferencing;
this._pRegular = pRegular;
this._qRegular = qRegular;
this._pSeason = pSeason;
this._qSeason = qSeason;
this._seasonDifferencingLevel = seassonDifferencing;
this._seasonPartern = seasonPartern;
Statistic.ComputeDifference(ref _processARIMASeries, ref _startIndex, _regularDifferencingLevel, _seasonDifferencingLevel, _seasonPartern);
EstimateARIMACoef(_processARIMASeries, _startIndex, _pRegular, _qRegular, _seasonPartern, _pSeason, _qSeason, out _listArimaCoef);
List <double> trainingResultSeries = new List <double>();
ComputeTrainingResultData(_processARIMASeries, 0, _regularDifferencingLevel, _seasonDifferencingLevel, _pRegular, _qRegular, _seasonPartern, _pSeason, _qSeason, _listArimaCoef, out trainingResultSeries);
List <double> currentARIMASeries = _processARIMASeries.FindAll(item => true);
int startIndexTraining = _startIndex;
RevertDiffTestSeries(ref currentARIMASeries, ref trainingResultSeries, ref startIndexTraining, _regularDifferencingLevel, _seasonDifferencingLevel, _seasonPartern);
_errorARIMASeries = new List <double>();
for (int i = 0; i < _originARIMASeries.Count; i++)
{
_errorARIMASeries.Add(_originARIMASeries[i] - trainingResultSeries[i]);
}
//ComputTrainingError(_processARIMASeries, _startIndex, _regularDifferencingLevel, _seasonDifferencingLevel, _pRegular, _qRegular, _seasonPartern, _pSeason, _qSeason, _listArimaCoef, out _errorARIMASeries);
}
private void ForecastARIMA(List <double> timeSeries, int regularDifferencingLevel, int seasonDifferencingLevel, int pRegular, int qRegular, int seasonPartern, int pSeason, int qSeason, List <double> listArimaCoeff, int nHead, out List <double> forecastSeries)
{
List <double> currentSeries = timeSeries.FindAll(item => true);
List <double> currentErrors = new List <double>();
List <double> currentResults = new List <double>();
int startIndexForecast = 0;
Statistic.ComputeDifference(ref currentSeries, ref startIndexForecast, _regularDifferencingLevel, _seasonDifferencingLevel, _seasonPartern);
int begin = ComputeMax(pRegular, qRegular, pSeason * seasonPartern, qSeason * seasonPartern, startIndexForecast);
for (int i = 0; i < begin; i++)
{
currentResults.Add(currentSeries[i]);
currentErrors.Add(0);
}
for (int i = begin; i < currentSeries.Count; i++)
{
double temp = listArimaCoeff[0];
for (int j = 1; j <= pRegular; j++)
{
temp += currentSeries[i - j] * listArimaCoeff[j];
}
for (int j = 1; j <= pSeason; j++)
{
temp += currentSeries[i - j * seasonPartern] * listArimaCoeff[pRegular + qRegular + j];
}
currentResults.Add(temp);
currentErrors.Add(timeSeries[i] - currentSeries[i]);
}
for (int i = 0; i < nHead; i++)
{
double temp = listArimaCoeff[0];
for (int j = 1; j <= pRegular; j++)
{
temp += currentSeries[timeSeries.Count + i - j] * listArimaCoeff[j];
}
for (int j = 1; j <= pSeason; j++)
{
temp += currentSeries[timeSeries.Count + i - j * seasonPartern] * listArimaCoeff[pRegular + qRegular + j];
}
currentResults.Add(temp);
currentSeries.Add(temp);
currentErrors.Add(timeSeries[i] - currentSeries[i]);
}
RevertDiffTestSeries(ref currentSeries, ref currentResults, ref startIndexForecast, regularDifferencingLevel, seasonDifferencingLevel, seasonPartern);
forecastSeries = new List <double>();
for (int i = 0; i < nHead; i++)
{
forecastSeries.Add(currentSeries[timeSeries.Count + i]);
}
}
public void RemoveTrendSeasonality(int regularDifferencingLevel, int seasonDifferencingLevel, int seasonPartern)
{
InitTraining();
_regularDifferencingLevel = regularDifferencingLevel;
_seasonDifferencingLevel = seasonDifferencingLevel;
_seasonPartern = seasonPartern;
Statistic.ComputeDifference(ref _processARIMASeries, ref _startIndex, _regularDifferencingLevel, _seasonDifferencingLevel, _seasonPartern);
}
public void ComputeTestingResult(List <double> testDataSeries, out List <double> testResultSeries)
{
List <double> testDataSeriesProcessed = testDataSeries.FindAll(item => true);
testResultSeries = new List <double>();
int startIndexTest = 0;
Statistic.ComputeDifference(ref testDataSeriesProcessed, ref startIndexTest, _regularDifferencingLevel, _seasonDifferencingLevel, _seasonPartern);
ComputeTrainingResultData(testDataSeriesProcessed, startIndexTest, _regularDifferencingLevel, _seasonDifferencingLevel, _pRegular, _qRegular, _seasonPartern, _pSeason, _qSeason, _listArimaCoef, out testResultSeries);
RevertDiffTestSeries(ref testDataSeriesProcessed, ref testResultSeries, ref startIndexTest, _regularDifferencingLevel, _seasonDifferencingLevel, _seasonPartern);
}