DifferencingTransformer<T>
Models & Types · AiDotNet.Preprocessing.TimeSeries
Applies differencing and stationarity transformations to time series data.
Inheritance: TimeSeriesTransformerBase<T> → DifferencingTransformer<T>
For Beginners
Many forecasting methods assume your data doesn’t have trends or seasonal patterns. This transformer removes those patterns through differencing.
Common transforms include:
- First Difference: Change from previous value (removes linear trend)
- Seasonal Difference: Change from same time in previous season (removes seasonality)
- Detrending: Subtract a fitted trend line
- Decomposition: Separate trend, seasonal, and residual components
After transformation, you can check stationarity using statistical tests like the Augmented Dickey-Fuller test.
How It Works
This transformer provides various methods to make time series stationary, which is a requirement for many forecasting models (ARIMA, VAR, etc.). Stationary data has constant statistical properties over time.
Constructors
| Constructor | Summary |
|---|---|
DifferencingTransformer(TimeSeriesFeatureOptions options) |
Creates a new differencing transformer with the specified options. |
Properties
| Property | Summary |
|---|---|
bool SupportsInverseTransform { get; } |