PsiForecastETS(target_date, values, timeline, [, seasonality][, data_completion][, aggregation][, simulate])

PsiForecastETS() uses exponential smoothing to predict a future value or values in a time series dataset with the option to either detect seasonality in the dataset automatically or pass a seasonality period.  Passing False as the last argument will result in a static forecast.  If True is passed for this argument, a random error will be included in the forecasted points. For more information on performing a time series forecast using these two functions, please see the Analytic Solver User Guide section, “Excel 2016 Forecast Functions" in the chapter, “Getting Results:  Simulation.”

target_date: A data point for which you want the predicted value. A data point may be date/time or numeric. If a target date is given that appears before the start of the timeline in the dataset, PsiForecastETS will return #NUM (i.e. if, in this example, 1940 is passed for X)

values: An Excel range containing the historical values in the given dataset.

timeline: An Excel range containing the time variables in the given dataset. Both functions require the historical data to be structured using a constant interval between data points. For example, the Airpass example dataset (Help -- Example Models -- Forecasting\Data Mining Examples) presents monthly passenger data, and therefore, the forecast must also predict the number of passengers by month.

seasonality: (Optional) This argument indicates the length of the seasonal pattern. The following values are accepted as valid inputs.  All other values will return a #NUM error. 

0:  Signifies no seasonality exists in the data.  The result is a linear prediction. 

1: (Default Value) Triggers Solver to detect seasonality within the data automatically.  

1 < N < 8,760:  Positive integer values greater than 1 but less will be used as the seasonality period. 

data_completion:   (Optional) This argument specifies  how to handle missing values.  The default value of 1 replaces missing values by interpolation.  If 0 is passed, missing values will be replaced by 0's. 

aggregation:  (Optional)  PsiForecastETS can aggregate multiple points with the same time stamp.  Pass an integer value from 0 to 6 to indicate which method should be used.

0:  (Default)  Average

1:  SUM



4:  MIN

5:  MAX


simulate: (Optional) Pass True or False for the third argument.  Passing False (the default) will result in a static forecast that will only update if a cell passed in the 2nd argument is changed.  If True is passed for this argument, a random error will be included in the forecasted points. 

Note:  This function was introduced to coincide with the Excel 2016 Forecast function, Forecast.ETS. This function computes a forecast using exponential smoothing, similar to the PsiForecast function, but uses different argument lists and also differs in other specific details.  Analytic Solver (starting with V2018) includes support for this function, so you can use it in your models, with one caveat:  If you use this forecast function in an optimization model where the arguments depend on the decision variables, or in a simulation model where the arguments depend on uncertain variables, you may see slightly different calculated forecast values depending on whether the Psi Interpreter or Excel Interpreter is selected in the Task Pane Platform tab.  When the Psi Interpreter is used, a slightly different (but arguably better) methodology is used to compute these forecast functions.  When the Excel Interpreter is selected, Microsoft Excel is used to calculate the worksheet and thus Excel's methodology will be used.