AIssisted™ Time Series Prediction Preview


When you make a forecast, the AIssisted™ Time Series Prediction Wizard takes your source data and calculates prediction data. The AIssisted™ Prediction Preview then shows these two data types for comparison, complete with table, graph, and accuracy percentage.

Once you have defined a particular area of a cube using the wizard, the Prediction Preview shows the figures of the data populating this area. Choose the Measure and Comparison version from the combo boxes to compare them with the prediction data.

The below example shows the predicted forecast for Gross Revenue, with Actual chosen as the comparison dimension:

You see the predicted values along with their upper and lower bound values as lines, with the actual and comparison versions shown as bars. You can display different data by choosing a different Measure or Comparison version from the combo boxes. If you select a measure other than Actual, the lighter blue bar will change, depicting the comparison measure. (You can hide a line in the graph by clicking on the corresponding, color-coordinated labels below the graph.)

Here is an example of a prediction forecast for the months of 2019-07 to 2020-06 with the Comparison version set to Budget:

The fields and functionality of the Prediction Preview are described below:

Field Description
Measure (combo box) Choose the Measure element you wish to use for your predictive forecast (e.g. Gross Revenue). You can select different elements to show different prediction results.
Comparison version (combo box) Choose the Comparison version element you wish to use for comparing with your prediction data (e.g. Actual). It is possible to select different elements to compare with the predicted data; for instance, you can compare the accuracy of prediction of a given measure from Actual to the Budget or other. 
Reset (button) Resets the duration of the prediction to your original specification (e.g. 6 months, 01-06, 2018).
-3 months (button)

Increases the duration of the prediction by adding three previous months of data (e.g. 09-12. 2017). The -3 months button creates a larger number of data points on the graph, showing the previous three months of data for comparison with the historical data (e.g. 09-12. 2017). If you have additional prediction data in the cube for these dates, this will likely alter the accuracy of your prediction.

Actual (source) The actual figures for the duration of your prediction. It is what your prediction compares itself with to determine its accuracy. The source is always the data you entered from the Version dimension, e.g. Actual, and will always be expressed in the table and graph.
[Actual] (comparison) This is a dynamic figure that compares to the predicted forecast. Note: a change in the Comparison version element will be reflected in the table and graph; that is, if you switch it from Actual to Budget, the field in the table will become Budget (comparison), and Budget will be shown in the graph.
Prediction The forecast data produced by the AIssisted™ Time Series Prediction Wizard.
Prediction_lower The lower bounds of your prediction data, i.e. the lowest point your data is expected to sink below the predicted forecast.
Prediction_upper The upper bounds of your prediction data, i.e. the highest point your data is expected to rise above the predicted forecast.
Accuracy The relative accuracy of your prediction and its Comparison version expressed as a percentage (e.g. the prediction above has an accuracy of 96.18% compared to the Budget data above).

Note: although you can make a predictive forecast based on fewer than 36 months of data, we strongly suggest 36 months as a minimum for optimal accuracy.