/// <summary>
/// Aan de hand van de ingevulde kolommen van ImproveSmoothingFactor de prediction berekenen.
/// </summary>
/// <param name="data">Lijst van Maand waarbij de forecasting data wordt opgeslagen</param>
private void AddOptimizePrediction(Dictionary <int, Forecasting> data)
{
var preditionValue = new PredictionVariables();
var predictionCalc = new SmoothingFormulas();
var factorValues = new FactorsVariables();
factorValues.alpha = bestAlpha;
factorValues.delta = bestBeta;
factorValues.gamma = bestGamma;
foreach (var item in data)
{
int timeValue = item.Key;
if (timeValue == 36)
{
preditionValue.tLock = item.Key;
preditionValue.tTLock = item.Value.Trend;
preditionValue.lTLock = item.Value.Level;
}
else if (timeValue > 36)
{
preditionValue.t = timeValue;
preditionValue.sT = data[int.Parse(timeValue.ToString()) - 12].SeasonalAdjustment;
Console.WriteLine("Month: " + timeValue + ", " + "Predicted Demand Value: " + predictionCalc.Prediction(preditionValue, factorValues));
data[int.Parse(timeValue.ToString())].ActualDemand = predictionCalc.Prediction(preditionValue, factorValues);
}
}
}
public float Prediction(PredictionVariables predictionVariables, FactorsVariables factorsVariables)
{
//(lT - 1 + (t.current - t - 1) * Tt - 1) * S(vorige 12 maanden);
//---->($LevelsmoothingBekend + (PredictionT - $lastbekendeTijd) * $LastbekendeTrend) *-12Seasonal
float prediction = (predictionVariables.lTLock + (predictionVariables.t - predictionVariables.tLock) * predictionVariables.tTLock) * predictionVariables.sT;
return(prediction);
}
/// <summary>
/// Vullen kolommen van Forecasting volgens de formules
/// </summary>
/// <param name="data">Lijst van Maand waarbij de forecasting data wordt opgeslagen</param>
private void ImproveSmoothingFactor(Dictionary <int, Forecasting> data)
{
var formulas = new SmoothingFormulas();
FactorsVariables newFactorsVariables = new FactorsVariables();
newFactorsVariables.alpha = alpha;
newFactorsVariables.delta = beta;
newFactorsVariables.gamma = gamma;
SmoothingVariables newSmoothingVariables = new SmoothingVariables();
int key = -11;
foreach (var item in data)
{
if (item.Key == 1)
{
//magic sequence
item.Value.OneStepForecast = formulas.OneStepErrorCalc(data[0].Level, data[0].Trend, data[key].SeasonalAdjustment);
item.Value.ForecastError = formulas.ForecastErrorCalc(item.Value.ActualDemand, item.Value.OneStepForecast);
item.Value.SquaredError = formulas.SquardErrorCalc(item.Value.ForecastError);// debug
newSmoothingVariables.forecastError = item.Value.ForecastError;
newSmoothingVariables.oneStepForecastError = item.Value.OneStepForecast;
newSmoothingVariables.xT = data[item.Key - 1].ActualDemand;
newSmoothingVariables.lT = data[item.Key - 1].Level;
newSmoothingVariables.sT = data[key].SeasonalAdjustment;
newSmoothingVariables.tT = data[item.Key - 1].Trend;
//TODO make formulas simple
newSmoothingVariables.lT = item.Value.Level = formulas.SmoothLevelCalc(newSmoothingVariables, newFactorsVariables);
newSmoothingVariables.tT = item.Value.Trend = formulas.SmoothTrendCalc(newSmoothingVariables, newFactorsVariables);
newSmoothingVariables.sT = item.Value.SeasonalAdjustment = formulas.SmoothSeasonalCalc(newSmoothingVariables, newFactorsVariables);
key++;
}
else if (item.Key > 1 && item.Key <= 36)
{
//magic all
//Paint
item.Value.OneStepForecast = formulas.OneStepErrorCalc(data[item.Key - 1].Level, data[item.Key - 1].Trend, data[key].SeasonalAdjustment);
item.Value.ForecastError = formulas.ForecastErrorCalc(item.Value.ActualDemand, item.Value.OneStepForecast);
item.Value.SquaredError = formulas.SquardErrorCalc(item.Value.ForecastError);
newSmoothingVariables.forecastError = item.Value.ForecastError;
newSmoothingVariables.oneStepForecastError = item.Value.OneStepForecast;
newSmoothingVariables.xT = data[item.Key - 1].ActualDemand;
newSmoothingVariables.lT = data[item.Key - 1].Level;
newSmoothingVariables.sT = data[key].SeasonalAdjustment;
newSmoothingVariables.tT = data[item.Key - 1].Trend;
//TODO make formulas simple
newSmoothingVariables.lT = item.Value.Level = formulas.SmoothLevelCalc(newSmoothingVariables, newFactorsVariables);
newSmoothingVariables.tT = item.Value.Trend = formulas.SmoothTrendCalc(newSmoothingVariables, newFactorsVariables);
newSmoothingVariables.sT = item.Value.SeasonalAdjustment = formulas.SmoothSeasonalCalc(newSmoothingVariables, newFactorsVariables);
key++;
}
}
}
public float SmoothSeasonalCalc(SmoothingVariables seasonalVariables, FactorsVariables factorsVariables)
{
// Ct = sT + delta * (1 - alpha ) * ForecastError /(lT-1 + tT-1)
// seasonalVariables.oneStepForecastError = OneStepErrorCalc(seasonalVariables.lT,seasonalVariables.tT,seasonalVariables.sT);
// seasonalVariables.forecastError = ForecastErrorCalc(seasonalVariables.xT, seasonalVariables.oneStepForecastError);
float sT = seasonalVariables.sT + factorsVariables.delta * (1 - factorsVariables.alpha) * seasonalVariables.forecastError / (seasonalVariables.lT + seasonalVariables.tT);
return(sT);
}
public float SmoothTrendCalc(SmoothingVariables trendVariables, FactorsVariables factorsVariables)
{
// Bt = bt-1 + gemma * alpha * (forecastError/ct)
//trendVariables.oneStepForecastError = OneStepErrorCalc(trendVariables.lT, trendVariables.tT, trendVariables.sT);
//trendVariables.forecastError = ForecastErrorCalc(trendVariables.xT, trendVariables.oneStepForecastError);
float tT = trendVariables.tT + factorsVariables.gamma * factorsVariables.alpha * (trendVariables.forecastError / trendVariables.sT);
return(tT);
}
public float SmoothLevelCalc(SmoothingVariables levelVariables, FactorsVariables factorsVariables)
{
//lT-1 + Bt-1 + alpha * (forecastError/Ct)
//levelVariables.oneStepForecastError = OneStepErrorCalc(levelVariables.lT, levelVariables.tT, levelVariables.sT);
//levelVariables.forecastError = ForecastErrorCalc(levelVariables.xT, levelVariables.oneStepForecastError);
float lT = levelVariables.lT + levelVariables.tT + factorsVariables.alpha * (levelVariables.forecastError / levelVariables.sT);
return(lT);
}
