Creating a multivariate forecast

Edit online

Use multivariate forecasting to include external variables and improve prediction accuracy.

Before you begin

Ensure that you have access to the Baseline Forecast worksheet and the Forecast cube data for the main and external variables.

About this task

Multivariate forecasting includes external variables, called exogenous factors, such as production volume, headcount, or temperature. The Planning Analytics Workspace assistant analyzes the main variable and the selected external variables to build a predictive model.

Tip:
  • Select variables that logically influence your target variable. For example, production volume affects energy consumption.
  • Verify that variable data covers both historical and future periods.
  • Do not use seasonality-based spreading for multivariate forecasts. Use the last historical period method instead.
  • Validate model performance in the Statistical Details tab in the forecast preview.

Procedure

  1. Open the Baseline Forecast worksheet and identify the main variable to forecast, such as Electricity.
  2. Select the row for the location to forecast, such as Grand Central.
  3. From the top-level menu, select Forecast > Multivariate Forecast.
  4. In the side panel, select Add variables and choose the Forecast cube.
  5. Specify the same location and utility type as the main variable.
    For example, Grand Central, Electricity(kWh), Native Unit, and so on.
  6. Set Version to Actuals + Forecast. Select Include future values to allow future (exogenous) values to be used as predictors for predicted time periods.
  7. In the Advanced section, confirm Dimension = Version, Hierarchy = Version, Member to save prediction = Forecast and Spreading = Last historical period.
  8. Select Preview to review the model and variable contributions.
  9. Select Forecast to save results to the Forecast version.
  10. Repeat the process for the entire portfolio to create a BAU scenario for the entire organization.

Results

This forecast includes the main variable and the selected external variables for improved accuracy.

What to do next

After you create a baseline forecast, create target scenarios for an ideal future emissions pathway.