Creating a multivariate forecast

A multivariate forecast is a forecast that takes into consideration more than one time-dependent variable. The forecast model not only looks at historical data but also the dependencies between the specified variables to create a forecast.

In IBM® Cognos Analytics with Watson, you can create a multivariate forecast that considers multiple measures in its model. To create a multivariate forecast:
  1. Create a line, bar, or a column visualization.
  2. Enable Forecast under the Analytics panel.

    Turn on Forecast
  3. Click the Optional factors to consider slot and select the measures you want the forecast to consider. You can also drag measures from the data tree to this slot.

    Note: You can add a maximum of five measures for consideration.
    Optional factors to consider

The optional factors are not displayed in the forecast. Hover over the forecasted values to see the factors that were considered.

If the forecast determines that the measures you selected are independent of each other, it displays a univariate model instead. A univariate model creates a forecast for one variable without considering any other variables.

To see which forecast model was chosen for forecasting, click See statistical details and check the Notes column under Forecasting statistical details. You can also see which multivariate model was used – ARIMAX or Vector-Autoregression.

If you used the Fields tab to generate a forecast, you can also add a column to the Color slot (group by that measure). When you populate the Color slot in the visualization, any measure that you add to Optional factors to consider also applies to the Color slot.

Note: When you add a filtered measure on the Color slot, the optional factors that are taken into consideration apply only to the filtered options and not to any other options in the measure.

The following example shows Box office average by Year that is grouped by Story (in the Color slot). Story is also filtered to only display Comedy, Discovery, and Love.

Visualization showing Analytics panel

When you add an optional factor to consider, the forecast considers Opening Weekend for only the filtered options in Story (Opening Weekend for Comedy, Opening weekend for Discovery, and Opening Weekend for Love). The optional factor is not considered for other options in the Story measure.

Multivariate forecast with Color slot