DifferencingFeatures
Enums · AiDotNet.Models.Options
Flags for selecting which differencing and stationarity features to compute.
For Beginners
Differencing transforms help make time series stationary, which is required by many forecasting models. A stationary series has:
- Constant mean over time (no trend)
- Constant variance over time (no changing volatility)
- No seasonal patterns
Fields
| Field | Summary |
|---|---|
DifferencingFeatures All |
All available differencing features. |
DifferencingFeatures BasicDifferencing |
All basic differencing methods. |
DifferencingFeatures Decomposition |
All decomposition methods. |
DifferencingFeatures Detrending |
All detrending methods. |
DifferencingFeatures FirstDifference |
First-order differencing: y[t] - y[t-1]. |
DifferencingFeatures HodrickPrescottFilter |
Hodrick-Prescott filter: extracts trend and cycle components. |
DifferencingFeatures LinearDetrend |
Linear detrending: removes best-fit straight line. |
DifferencingFeatures LogDifference |
Log difference: log(y[t]) - log(y[t-1]). |
DifferencingFeatures None |
No differencing features. |
DifferencingFeatures PercentChange |
Percent change: (y[t] - y[t-1]) / y[t-1]. |
DifferencingFeatures PolynomialDetrend |
Polynomial detrending: removes best-fit polynomial. |
DifferencingFeatures Returns |
All return-based transforms. |
DifferencingFeatures SeasonalDifference |
Seasonal differencing: y[t] - y[t-period]. |
DifferencingFeatures SecondDifference |
Second-order differencing: diff(diff(y)). |
DifferencingFeatures StlDecomposition |
STL decomposition: Seasonal-Trend decomposition using LOESS. |
int value__ |