Running a Forecast in a Time Series or Line Chart

You can run a predictive forecast in time series or line charts.

Context

You can use a time series or line chart to automatically predict future data values based on the data included in your chart definition. The creation of a forecast requires a measure and a date dimension, along with additional options that can be integrated into the visualization:
  • Multiple measures can be used to compare predicted values across these measures.
  • Coloring by dimensions can be used to compare predicted values across dimension values.
Once data is loaded into the chart, you can perform any of the following forecasting options:
  • Automatic Forecast: performs a predictive forecast on the available data. With this option, you can specify how many forecast periods to display in the chart.
  • Advanced Options: enables you to select one of the following simulate forecasting scenarios:
    • Linear Regression: use the Linear Regression algorithm on historical time series data to predict future values.
    • Triple Exponential Smoothing: use the Triple Exponential Smoothing algorithm to account for seasonal changes as well as trends.
    • Add Additional Inputs: use calculated measures, measure input controls, or additional measures from the current model as inputs when creating the forecast. This option is available only for Time Series charts.
      Note
      To avoid errors when using Additional Inputs, ensure that data is available for both your original input data periods as well as the future period(s) you would like to forecast.
Note
Before you can create a forecast on a time series chart, using a remote model from an SAP HANA, SAP BW, SAP Universe or SAP S/4HANA system, your administrator must configure the following option: Enable Time Series Forecast and Smart Grouping on Live Data Models. The data will be processed on SAP servers. You cannot run forecasts on Line charts using remote models.
The chart will display the following information, either in the chart or in the tooltip:
  • Predicted values
  • Upper and lower limits for the confidence interval

Procedure

  1. From Start of the navigation path Builder Next navigation step Chart StructureEnd of the navigation path, select Start of the navigation pathTrend Next navigation step Time SeriesEnd of the navigation path or Line.
  2. Add one or more measures.
  3. Add the Date dimension.
  4. Add a Color dimension.
  5. From your chart action menu () choose Start of the navigation path Add Next navigation step ForecastEnd of the navigation path, and then select one of the following:
    • Automatic Forecast

      Once the forecast is performed, a color-coded Forecast link is displayed above the chart. The link is blue if the forecast was performed successfully with no data issues. The link is gray if the forecast quality is low, and red if the forecast was unsuccessful. Select the link to display a Forecast Periods slider for the number of displayed periods and the Forecast Quality for the configured data.

    • Advanced Options

      Choose one of the following advanced options for running your forecast:

      • Triple Exponential Smoothing

        Select this option to run the forecast using the Triple Exponential Smoothing algorithm to account for seasonal changes as well as trends. Once the forecast is performed, a color-coded Forecast link is displayed above the chart. The link is blue if the forecast was performed successfully with no data issues. The link is orange if the forecast was partially successful, and red if the forecast was unsuccessful. Select the link to display a Forecast Periods slider for the number of displayed periods and the Forecast Quality for the configured data.

      • Add Additional Inputs...
        Note
        This option is not available for Line charts.

        The Additional Forecast Input panel opens on the right.

      • Use the Forecast Periods slider to specify how many forecast periods to display in the chart.
      • Select +Add Additional Input to add any of the following additional inputs from the model used to create the chart:
        • Calculated measures
        • Additional measures
        • A measure input control

Results

The resulting chart displays the forecast values and the confidence interval for the forecast. To show or hide the confidence interval, select (More Actions)Start of the navigation path Next navigation step Show/Hide Next navigation step Confidence IntervalEnd of the navigation path.

To see how well the predictive model matches with the historical data, select (More Actions)Start of the navigation path Next navigation step Show/Hide Next navigation step Past Period ForecastEnd of the navigation path. A dashed line (originating from the first point of the forecast) depicts the predictive model applied to past periods.

Related Information