Time Series node - model options

Model name. You can generate the model name automatically based on the target or ID field (or model type in cases where no such field is specified) or specify a custom name.

Confidence limit width (%). Confidence intervals are computed for the model predictions and residual autocorrelations. You can specify any positive value less than 100. By default, a 95% confidence interval is used.

Continue estimation using existing model(s). If you already generated a time series model, select this option to reuse the criteria settings that are specified for that model and generate a new model node in the Models palette, rather than building a new model from the beginning. In this way, you can save time by re-estimating and producing a new forecast that is based on the same model settings as before but using more recent data. Thus, for example, if the original model for a particular time series was Holt's linear trend, the same type of model is used for reestimating and forecasting for that data. The system does not reattempt to find the best model type for the new data.

Build scoring model only. To reduce the amount of data that is stored in the model, check this box. Using this option can improve performance when building models with large numbers of time series (tens of thousands). You can still score the data in the usual way.

Extend records into the future. Enables the following Future Values to Use in Forecasting section, where you can set the number of time intervals to forecast beyond the end of the estimation period. The time interval in this case is the time interval of the analysis, which you specify on the Data Specifications tab. There is no maximum limit for this setting. Using the following options, you can automatically compute future values of inputs, or you can manually specify forecasting values for one or more predictors.

Future Values to Use in Forecasting

Compute future values of inputs If you select this option, the forecast values for predictors, noise predictions, variance estimation, and future time values are calculated automatically. When forecasts are requested, autoregressive models are automatically built for any input series that are not also targets. These models are then used to generate values for those input series in the forecast period.

Select fields whose values you wish to add to the data. For each record that you want to forecast (excluding holdouts), if you are using predictor fields (with the role set to Input), you can specify estimated values for the forecast period for each predictor. You can either specify values manually, or choose from a list.

Field. Click the field selector button and choose any fields that may be used as predictors. Note that fields selected here may or may not be used in modeling; to actually use a field as a predictor, it must be selected in a downstream modeling node. This dialog box simply gives you a convenient place to specify future values so they can be shared by multiple downstream modeling nodes without specifying them separately in each node. Also note that the list of available fields may be constrained by selections on the Build Options tab.
Note that if future values are specified for a field that is no longer available in the stream (because it has been dropped or because of updated selections made on the Build Options tab), the field is shown in red.

Values. For each field, you can choose from a list of functions, or click Specify to either enter values manually or choose from a list of predefined values. If the predictor fields relate to items that are under your control, or which are otherwise knowable in advance, you should enter values manually. For example, if you are forecasting next month's revenues for a hotel based on the number of room reservations, you could specify the number of reservations you actually have for that period. Conversely, if a predictor field relates to something outside your control, such as a stock price, you could use a function such as the most recent value or the mean of recent points.

The available functions depend on the measurement level of the field.

Table 1. Functions available for measurement levels
Measurement level	Functions
Continuous or Nominal field	Blank Mean of recent points Most recent value Specify
Flag field	Blank Most recent value True False Specify

Mean of recent points calculates the future value from the mean of the last three data points.

Most recent value sets the future value to that of the most recent data point.

True/False sets the future value of a flag field to True or False as specified.

Specify opens a dialog box for specifying future values manually, or choosing them from a predefined list.

Make Available for Scoring

You can set the default values here for the scoring options that appear on the dialog box for the model nugget.

Calculate upper and lower confidence limits. If selected, this option creates new fields (with the default prefixes $TSLCI- and $TSUCI-) for the lower and upper confidence intervals, for each target field.
Calculate noise residuals. If selected, this option creates a new field (with the default prefix $TSResidual-) for the model residuals for each target field, together with a total of these values.

Model Settings

Maximum number of models to be displayed in output. Specify the maximum number of models you want to include in the output. Note that if the number of models built exceeds this threshold, the models are not shown in the output but they're still available for scoring. Default value is 10. Displaying a large number of models may result in poor performance or instability.