Streaming Time Series node - general build options

Expert Modeler

Under Model Type, select the type of models you want to build:

• All models. The Expert Modeler considers both ARIMA and exponential smoothing models.
• Exponential smoothing models only. The Expert Modeler considers only exponential smoothing models.
• ARIMA models only. The Expert Modeler considers only ARIMA models.

Expert Modeler considers seasonal models. This option is only enabled if a periodicity is defined for the active dataset. When this option is selected, the Expert Modeler considers both seasonal and nonseasonal models. If this option is not selected, the Expert Modeler considers only nonseasonal models.

Expert Modeler considers sophisticated exponential smoothing models. When this option is selected, the Expert Modeler searches a total of 13 exponential smoothing models (7 of them existed in the original Time Series node, and 6 of them were added in version 18.1). If this option is not selected, the Expert Modeler only searches the original 7 exponential smoothing models.

Under Outliers, select from the following options

Detect outliers automatically. By default, automatic detection of outliers is not performed. Select this option to perform automatic detection of outliers, then select the desired outlier types.

For more information, see Outliers.

Input fields must have a measurement level of Flag, Nominal, or Ordinal and must be numeric (for example, 1/0, not True/False, for a flag field), before they are included in this list.

For more information, see Pulses and steps.

The Expert Modeler considers only simple regression and not arbitrary transfer functions for inputs that are identified as event or intervention fields on the Fields tab.

Exponential Smoothing

Model Type. Exponential smoothing models are classified as either seasonal or nonseasonal.¹ Seasonal models are only available if the periodicity defined by using the Time Intervals pane on the Data Specifications tab is seasonal. The seasonal periodicities are: cyclic periods, years, quarters, months, days per week, hours per day, minutes per day, and seconds per day. The following model types are available:

• Simple. This model is appropriate for a series in which there is no trend or seasonality. Its only relevant smoothing parameter is level. Simple exponential smoothing is most similar to an ARIMA with zero orders of autoregression, one order of differencing, one order of moving average, and no constant.
• Holt's linear trend. This model is appropriate for a series in which there is a linear trend and no seasonality. Its relevant smoothing parameters are level and trend, and, in this model, they are not constrained by each other's values. Holt's model is more general than Brown's model but may take longer to compute estimates for large series. Holt's exponential smoothing is most similar to an ARIMA with zero orders of autoregression, two orders of differencing, and two orders of moving average.
• Damped trend. This model is appropriate for a series in which there is a linear trend that is dying out and no seasonality. Its relevant smoothing parameters are level, trend, and damping trend. Damped exponential smoothing is most similar to an ARIMA with one order of autoregression, one order of differencing, and two orders of moving average.
• Multiplicative trend. This model is appropriate for a series in which there is a trend that changes with the magnitude of the series and no seasonality. Its relevant smoothing parameters are level and trend. Multiplicative trend exponential smoothing is not similar to any ARIMA model.
• Brown's linear trend. This model is appropriate for a series in which there is a linear trend and no seasonality. Its relevant smoothing parameters are level and trend, but, in this model, they are assumed to be equal. Brown's model is therefore a special case of Holt's model. Brown's exponential smoothing is most similar to an ARIMA with zero orders of autoregression, two orders of differencing, and two orders of moving average, with the coefficient for the second order of moving average equal to one half of the coefficient for the first order squared.
• Simple seasonal. This model is appropriate for a series in which there is no trend and a seasonal effect that is constant over time. Its relevant smoothing parameters are level and season. Seasonal exponential smoothing is most similar to an ARIMA with zero orders of autoregression; one order of differencing; one order of seasonal differencing; and orders 1, p, and p+1 of moving average, where p is the number of periods in a seasonal interval. For monthly data, p = 12.
• Winters' additive. This model is appropriate for a series in which there is a linear trend and a seasonal effect that is constant over time. Its relevant smoothing parameters are level, trend, and season. Winters' additive exponential smoothing is most similar to an ARIMA with zero orders of autoregression; one order of differencing; one order of seasonal differencing; and p+1 orders of moving average, where p is the number of periods in a seasonal interval. For monthly data, p=12.
• Damped trend with additive seasonal. This model is appropriate for a series in which there is a linear trend that is dying out and a seasonal effect that is constant over time. Its relevant smoothing parameters are level, trend, damping trend, and season. Damped trend and additive seasonal exponential smoothing is not similar to any ARIMA model.
• Multiplicative trend with additive seasonal. This model is appropriate for a series in which there is a trend that changes with the magnitude of the series and a seasonal effect that is constant over time. Its relevant smoothing parameters are level, trend, and season. Multiplicative trend and additive seasonal exponential smoothing is not similar to any ARIMA model.
• Multiplicative seasonal. This model is appropriate for a series in which there is no trend and a seasonal effect that changes with the magnitude of the series. Its relevant smoothing parameters are level and season. Multiplicative seasonal exponential smoothing is not similar to any ARIMA model.
• Winters' multiplicative. This model is appropriate for a series in which there is a linear trend and a seasonal effect that changes with the magnitude of the series. Its relevant smoothing parameters are level, trend, and season. Winters' multiplicative exponential smoothing is not similar to any ARIMA model.
• Damped trend with multiplicative seasonal. This model is appropriate for a series in which there is a linear trend that is dying out and a seasonal effect that changes with the magnitude of the series. Its relevant smoothing parameters are level, trend, damping trend, and season. Damped trend and multiplicative seasonal exponential smoothing is not similar to any ARIMA model.
• Multiplicative trend with multiplicative seasonal. This model is appropriate for a series in which there are a trend and a seasonal effect that both change with the magnitude of the series. Its relevant smoothing parameters are level, trend, and season. Multiplicative trend and multiplicative seasonal exponential smoothing is not similar to any ARIMA model.

Target Transformation. You can specify a transformation to be performed on each dependent variable before it is modeled.

For more information, see Series transformations.

• None. No transformation is performed.
• Square root. Square root transformation is performed.
• Natural log. Natural log transformation is performed.

ARIMA

Specify the structure of a custom ARIMA model.

ARIMA Orders. Enter values for the various ARIMA components of your model into the corresponding cells of the grid. All values must be non-negative integers. For autoregressive and moving average components, the value represents the maximum order. All positive lower orders are included in the model. For example, if you specify 2, the model includes orders 2 and 1. Cells in the Seasonal column are only enabled if a periodicity is defined for the active dataset.

• Autoregressive (p). The number of autoregressive orders in the model. Autoregressive orders specify which previous values from the series are used to predict current values. For example, an autoregressive order of 2 specifies that the value of the series two time periods in the past is used to predict the current value.
• Difference (d). Specifies the order of differencing applied to the series before estimating models. Differencing is necessary when trends are present (series with trends are typically nonstationary and ARIMA modeling assumes stationarity) and is used to remove their effect. The order of differencing corresponds to the degree of series trend; first-order differencing accounts for linear trends, second-order differencing accounts for quadratic trends, and so on.
• Moving Average (q). The number of moving average orders in the model. Moving average orders specify how deviations from the series mean for previous values are used to predict current values. For example, moving-average orders of 1 and 2 specify that deviations from the mean value of the series from each of the last two time periods be considered when predicting current values of the series.

Seasonal. Seasonal autoregressive, moving average, and differencing components play the same roles as their nonseasonal counterparts. For seasonal orders, however, current series values are affected by previous series values that are separated by one or more seasonal periods. For example, for monthly data (seasonal period of 12), a seasonal order of 1 means that the current series value is affected by the series value 12 periods before the current one. A seasonal order of 1, for monthly data, is then the same as specifying a nonseasonal order of 12.

Detect outliers automatically. Select this option to perform automatic detection of outliers, and select one or more of the outlier types available.

Type of Outliers to Detect. Select the outlier type(s) you want to detect. The supported types are:

• Additive (default)
• Level shift (default)
• Innovational
• Transient
• Seasonal additive
• Local trend
• Additive patch

Transfer Function Orders and Transformations. To specify transformations and to define transfer functions for any or all of the input fields in your ARIMA model, click Set; a separate dialog box is displayed in which you enter the transfer and transformation details.

Include constant in model. Inclusion of a constant is standard unless you are sure that the overall mean series value is 0. Excluding the constant is recommended when differencing is applied.

Further details

• For more information on types of outliers, see Outliers.
• For more information on transfer and transformation functiions, see Transfer and transformation functions.