Class DynamicRegressionWithARIMAErrorsOptions<T>
Configuration options for Dynamic Regression with ARIMA Errors, a powerful time series forecasting method that combines regression with time series error correction.
public class DynamicRegressionWithARIMAErrorsOptions<T> : TimeSeriesRegressionOptions<T>Type Parameters
TThe numeric type used for calculations (typically double or float).
- Inheritance
-
DynamicRegressionWithARIMAErrorsOptions<T>
- Inherited Members
Remarks
Dynamic Regression with ARIMA Errors (sometimes called RegARIMA) is a hybrid forecasting approach that uses external variables (regressors) to explain the main trend while modeling the residual errors with ARIMA (AutoRegressive Integrated Moving Average) to capture temporal patterns not explained by the regressors.
For Beginners: This is a forecasting method that combines two powerful techniques. First, it uses regression to find relationships between your target variable (what you're trying to predict) and other variables that might influence it (like temperature affecting ice cream sales). Then, it analyzes the errors in that prediction to find patterns over time (like seasonal trends or cycles). By combining both approaches, it often produces more accurate forecasts than either method alone. Think of it as not just predicting based on related factors, but also learning from its own mistakes.
Properties
AROrder
Gets or sets the AutoRegressive (AR) order for the ARIMA component.
public int AROrder { get; set; }Property Value
- int
The AR order, defaulting to 1.
Remarks
The AR order (p) specifies how many previous time steps of the error term are used to predict the current error. Higher values capture longer-term dependencies but increase model complexity.
For Beginners: This determines how far back in time the model looks when analyzing its errors. With the default value of 1, the model considers errors from one time step ago to help predict the current value. For example, if yesterday's forecast was too high, the model might adjust today's forecast downward. Higher values (like 2 or 3) would make the model consider patterns from multiple previous time periods.
DecompositionType
Gets or sets the matrix decomposition method used for solving the regression equations.
public MatrixDecompositionType DecompositionType { get; set; }Property Value
- MatrixDecompositionType
The matrix decomposition type, defaulting to LU decomposition.
Remarks
Different matrix decomposition methods offer trade-offs between numerical stability, computational efficiency, and accuracy. LU decomposition is a good general-purpose choice for many problems.
For Beginners: This is a technical setting that controls how the math behind the model is solved. The default (LU decomposition) works well for most situations. You typically won't need to change this unless you're dealing with special cases like highly correlated variables or numerical precision issues. Think of it as selecting which mathematical technique the computer uses to solve the equations in your model.
DifferenceOrder
Gets or sets the differencing order for the ARIMA component.
public int DifferenceOrder { get; set; }Property Value
- int
The differencing order, defaulting to 0.
Remarks
The differencing order (d) specifies how many times the data is differenced to achieve stationarity. Differencing helps remove trends and seasonality by computing the differences between consecutive observations.
For Beginners: This determines whether and how much the model transforms your data to remove upward or downward trends. With the default value of 0, no differencing is applied. A value of 1 means the model will work with the changes between consecutive values rather than the raw values themselves. This is useful when your data has a strong trend (like consistently increasing sales year over year). Think of it as focusing on the rate of change rather than the absolute values.
ExternalRegressors
Gets or sets the number of external regressor variables to use in the model.
public int ExternalRegressors { get; set; }Property Value
- int
The number of external regressors, defaulting to 1.
Remarks
External regressors are independent variables that help explain the behavior of the target variable. This setting determines how many such variables will be included in the regression component of the model.
For Beginners: This controls how many outside factors your model will consider when making predictions. With the default value of 1, the model will use one external factor (like temperature when predicting ice cream sales). If you have multiple factors that might influence your prediction (like temperature, day of week, and whether it's a holiday), you would increase this number accordingly.
MAOrder
Gets or sets the Moving Average (MA) order for the ARIMA component.
public int MAOrder { get; set; }Property Value
- int
The MA order, defaulting to 1.
Remarks
The MA order (q) specifies how many previous random shock terms are used in the prediction equation. It captures the short-term effects of random events on the time series.
For Beginners: This controls how the model handles unexpected events or "shocks" in your data. With the default value of 1, the model remembers the surprise factor from one time period ago. For example, if there was an unexpected spike in sales yesterday, the model accounts for how that surprise might affect today's prediction. Higher values make the model consider the lingering effects of surprises from multiple previous time periods.
Regularization
Gets or sets the regularization method used to prevent overfitting in the regression component.
public IRegularization<T, Matrix<T>, Vector<T>>? Regularization { get; set; }Property Value
- IRegularization<T, Matrix<T>, Vector<T>>
The regularization method, defaulting to null (no regularization).
Remarks
Regularization adds penalties to the model's complexity to prevent it from fitting noise in the training data. Common regularization methods include Ridge (L2), Lasso (L1), and ElasticNet (combination of L1 and L2).
For Beginners: This controls whether and how the model prevents itself from becoming too complex and "memorizing" your training data instead of learning general patterns. Without regularization (the default), the model might perform very well on your training data but poorly on new data. Adding regularization helps the model generalize better to new situations. Think of it as encouraging the model to find simpler explanations rather than complicated ones that might just be capturing noise in your data.