//
// Adaptive Rate Smoothing
//
public static ForecastTable adaptiveRateSmoothing(decimal[] values, int Extension, decimal MinGamma, decimal MaxGamma)
{
ForecastTable dt = new ForecastTable();
for (Int32 i = 0; i < (values.Length + Extension); i++)
{
//Insert a row for each value in set
DataRow row = dt.NewRow();
dt.Rows.Add(row);
row.BeginEdit();
//assign its sequence number
row["Instance"] = i;
if (i < values.Length)
{
row["Value"] = values[i];
if (i == 0)
{//initialize first row
row["Forecast"] = values[i];
}
else
{//calculate gamma and forecast value
DataRow priorRow = dt.Select("Instance=" + (i - 1).ToString())[0];
decimal PriorForecast = (Decimal)priorRow["Forecast"];
decimal PriorValue = (Decimal)priorRow["Value"];
//decimal Gamma = Math.Abs(TrackingSignal(dt, 3));
//if (Gamma < MinGamma)
// Gamma = MinGamma;
//if (Gamma > MaxGamma)
// Gamma = MaxGamma;
//row["Forecast"] = PriorForecast + (Gamma * (PriorValue - PriorForecast));
}
}
else
{//extension set, can't use actual values anymore
DataRow priorRow = dt.Select("Instance=" + (i - 1).ToString())[0];
decimal PriorForecast = (Decimal)priorRow["Forecast"];
decimal PriorValue = (Decimal)priorRow["Forecast"];
//decimal Gamma = Math.Abs(TrackingSignal(dt, 3));
//if (Gamma < MinGamma)
// Gamma = MinGamma;
//if (Gamma > MaxGamma)
// Gamma = MaxGamma;
//row["Forecast"] = PriorForecast + (Gamma * (PriorValue - PriorForecast));
}
row.EndEdit();
}
dt.AcceptChanges();
return(dt);
}
//
//Exponential Smoothing
//
// F(t+1) = ( Alpha * D(t) ) + (1 - Alpha) * F(t)
//
public static ForecastTable exponentialSmoothing(decimal[] values, int Extension, decimal Alpha)
{
ForecastTable dt = new ForecastTable();
for (Int32 i = 0; i < (values.Length + Extension); i++)
{
//Insert a row for each value in set
DataRow row = dt.NewRow();
dt.Rows.Add(row);
row.BeginEdit();
//assign its sequence number
row["Instance"] = i;
if (i < values.Length)
{//test set
row["Value"] = values[i];
if (i == 0)
{//initialize first value
row["Forecast"] = values[i];
}
else
{
//main area of forecasting
DataRow priorRow = dt.Select("Instance=" + (i - 1).ToString())[0];
decimal PriorForecast = (Decimal)priorRow["Forecast"];
decimal PriorValue = (Decimal)priorRow["Value"];
row["Forecast"] = PriorForecast + (Alpha * (PriorValue - PriorForecast));
}
}
else
{//extension has to use prior forecast instead of prior value
DataRow priorRow = dt.Select("Instance=" + (i - 1).ToString())[0];
decimal PriorForecast = (Decimal)priorRow["Forecast"];
decimal PriorValue = (Decimal)priorRow["Forecast"];
row["Forecast"] = PriorForecast + (Alpha * (PriorValue - PriorForecast));
}
row.EndEdit();
}
dt.AcceptChanges();
return(dt);
}
//CumulativeSignedError = Sum( E(t) )
public static decimal CumulativeSignedError(ForecastTable dt, int IgnoreInitial)
{
string Filter = "Error Is Not Null AND Instance > " + IgnoreInitial.ToString();
if (dt.Select(Filter).Length == 0)
{
return(0);
}
return((Decimal)dt.Compute("SUM(Error)", Filter));
}
//MeanAbsolutePercentError = Sum( |PercentError| ) / n
public static decimal MeanAbsolutePercentError(ForecastTable dt, int IgnoreInitial)
{
string Filter = "AbsolutePercentError Is Not Null AND Instance > " + IgnoreInitial.ToString();
if (dt.Select(Filter).Length == 0)
{
return(1);
}
return((Decimal)dt.Compute("AVG(AbsolutePercentError)", Filter));
}
//CumulativeAbsoluteError = Sum( |E(t)| )
public static decimal CumulativeAbsoluteError(ForecastTable dt, bool Holdout, int IgnoreInitial)
{
string Filter = "AbsoluteError Is Not Null AND Instance > " + IgnoreInitial.ToString();
if (Holdout)
{
Filter += " AND Holdout=True";
}
if (dt.Select(Filter).Length == 0)
{
return(0);
}
return((Decimal)dt.Compute("SUM(AbsoluteError)", Filter));
}
//MSE = Sum( E(t)^2 ) / n
public static decimal MeanSquaredError(ForecastTable dt, int IgnoreInitial, int DegreesOfFreedom)
{
decimal SquareError = 0;
string Filter = "Error Is Not Null AND Instance > " + IgnoreInitial.ToString();
DataRow[] rows = dt.Select(Filter);
if (rows.Length == 0)
{
return(0);
}
foreach (DataRow row in rows)
{
SquareError = (Decimal)Math.Pow(Double.Parse(row["Error"].ToString()), (Double)2.0);
}
return(SquareError / (dt.Rows.Count - DegreesOfFreedom));
}
//
//Simple Moving Average
//
// ( Dt + D(t-1) + D(t-2) + ... + D(t-n+1) )
// F(t+1) = -----------------------------------------
// n
public static ForecastTable simpleMovingAverage(decimal[] values, int Extension, int Periods, int Holdout)
{
ForecastTable dt = new ForecastTable();
for (Int32 i = 0; i < values.Length + Extension; i++)
{
//Insert a row for each value in set
DataRow row = dt.NewRow();
dt.Rows.Add(row);
row.BeginEdit();
//assign its sequence number
row["Instance"] = i;
if (i < values.Length)
{//processing values which actually occurred
row["Value"] = values[i];
}
//Indicate if this is a holdout row
if (i == 0)
{//Initialize first row with its own value
row["Forecast"] = values[i];
}
else if (i <= values.Length - Holdout)
{//processing values which actually occurred, but not in holdout set
decimal avg = 0;
DataRow[] rows = dt.Select("Instance>=" + (i - Periods).ToString() + " AND Instance < " + i.ToString(), "Instance");
foreach (DataRow priorRow in rows)
{
avg += (Decimal)priorRow["Value"];
}
avg /= rows.Length;
row["Forecast"] = avg;
}
else
{//must be in the holdout set or the extension
decimal avg = 0;
//get the Periods-prior rows and calculate an average actual value
DataRow[] rows = dt.Select("Instance>=" + (i - Periods).ToString() + " AND Instance < " + i.ToString(), "Instance");
foreach (DataRow priorRow in rows)
{
if ((Int32)priorRow["Instance"] < values.Length)
{//in the test or holdout set
avg += (Decimal)priorRow["Value"];
}
else
{//extension, use forecast since we don't have an actual value
avg += (Decimal)priorRow["Forecast"];
}
}
avg /= rows.Length;
//set the forecasted value
row["Forecast"] = avg;
}
row.EndEdit();
}
dt.AcceptChanges();
return(dt);
}
//
//Weighted Moving Average
//
// F(t+1) = (Weight1 * D(t)) + (Weight2 * D(t-1)) + (Weight3 * D(t-2)) + ... + (WeightN * D(t-n+1))
//
public static ForecastTable weightedMovingAverage(decimal[] values, int Extension, params decimal[] PeriodWeight)
{
//PeriodWeight[].Length is used to determine the number of periods over which to average
//PeriodWeight[x] is used to apply a weight to the prior period's value
//Make sure PeriodWeight values add up to 100%
decimal test = 0;
foreach (decimal weight in PeriodWeight)
{
test += weight;
}
if (test != 1)
{
return(null);
}
ForecastTable dt = new ForecastTable();
for (Int32 i = 0; i < values.Length + Extension; i++)
{
//Insert a row for each value in set
DataRow row = dt.NewRow();
dt.Rows.Add(row);
row.BeginEdit();
//assign its sequence number
row["Instance"] = i;
if (i < values.Length)
{//we're in the test set
row["Value"] = values[i];
}
if (i == 0)
{//initialize forecast with first row's value
row["Forecast"] = values[i];
}
else if ((i < values.Length) && (i < PeriodWeight.Length))
{//processing one of the first rows, before we've advanced enough to properly weight past rows
decimal avg = 0;
//Get the datarows representing the values within the WMA length
DataRow[] rows = dt.Select("Instance>=" + (i - PeriodWeight.Length).ToString() + " AND Instance < " + i.ToString(), "Instance");
for (int j = 0; j < rows.Length; j++)
{//apply an initial, uniform weight (1 / rows.Length) to the initial rows
avg += (Decimal)rows[j]["Value"] * (1 / rows.Length);
}
row["Forecast"] = avg;
}
else if ((i < values.Length) && (i >= PeriodWeight.Length))
{//Out of initial rows and processing the test set
decimal avg = 0;
//Get the rows within the weight range just prior to the current row
DataRow[] rows = dt.Select("Instance>=" + (i - PeriodWeight.Length).ToString() + " AND Instance < " + i.ToString(), "Instance");
for (int j = 0; j <= rows.Length - 1; j++)
{//Apply the appropriate period's weight to the value
avg += (Decimal)rows[j]["Value"] * PeriodWeight[j];
}
//Assign the forecasted value to the current row
row["Forecast"] = avg;
}
else
{//into the extension
decimal avg = 0;
DataRow[] rows = dt.Select("Instance>=" + (i - PeriodWeight.Length).ToString() + " AND Instance < " + i.ToString(), "Instance");
for (int j = 0; j < rows.Length; j++)
{//with no actual values to weight, use the previous rows' forecast instead
avg += (Decimal)rows[j]["Forecast"] * PeriodWeight[j];
}
row["Forecast"] = avg;
}
row.EndEdit();
}
dt.AcceptChanges();
return(dt);
}
