How to perform predictive forecasting in SAP Analytics Cloud

[box type="note" align="" class="" width=""]This article is an excerpt from a book written by Riaz Ahmed titled Learning SAP Analytics Cloud. This book involves features of the SAP Analytics Cloud which will help you collaborate, predict and solve business intelligence problems with cloud computing.[/box]

In this article we will learn how to use predictive forecasting with the help of a trend time series chart to see revenue trends in a range of a year.

Time series forecasting is only supported for planning models in SAP Analytics Cloud. So, you need planning rights and a planning license to run a predictive time-series forecast. However, you can add predictive forecast by creating a trend time series chart based on an analytical model to estimate future values. In this article, you will use a trend time series chart to view net revenue trends throughout the range of a year. A predictive time-series forecast runs an algorithm on historical data to predict future values for specific measures. For this type of chart, you can forecast a maximum of three different measures, and you have to specify the time for the prediction and the past time periods to use as historical data.

• Add a blank chart from the Insert toolbar.
• Set Data Source to the BestRun_Demo model.
• Select the Time Series chart from the Trend category.
• In the Measures section, click on the Add Measure link, and select Net Revenue.

Finally, click on the Add Dimension link in the Time section, and select Date as the chart’s dimension:

The output of your selections is depicted in the first view in the following screenshot. Every chart you create on your story page has its own unique elements that let you navigate and drill into details. The trend time series chart also allows you to zoom in to different time periods and scroll across the entire timeline. For example, the first figure in the following illustration provides a one-year view (A) of net revenue trends, that is from January to December 2015. Click on the six months link (B) to see the corresponding output, as illustrated in the second view. Drag the rectangle box (C) to the left or right to scroll across the entire timeline:

Adding a forecast

Click on the last data point representing December 2015, and select Add Forecast from the More Actions menu (D) to add a forecast:

You see the Predictive Forecast panel on the right side, which displays the maximum number of forecast periods. Using the slider (E) in this section, you can reduce the number of forecast periods. By default, you see the maximum number (in the current scenario, it is seven) in the slider, which is determined by the amount of historical data you have. In the Forecast On section, you see the measure (F) you selected for the chart. If required, you can forecast a maximum of three different measures in this type of chart that you can add in the Builder panel. For the time being, click on OK to accept the default values for the forecast, as illustrated in the following screenshot:

The forecast will be added to the chart. It is indicated by a highlighted area (G) and a dotted line (H). Click on the 1 year link (I) to see an output similar to the one illustrated in the following screenshot under the Modifying forecast section. As you can see, there are several data points that represent forecast. The top and bottom of the highlighted area indicate the upper and lower bounds of the prediction range, and the data points fall in the middle (on the dotted line) of the forecast range for each time period. Select a data point to see the Upper Confidence Bound (J) and Lower Confidence Bound (K) values.

Modifying forecast

You can modify a forecast using the link provided in the Forecast section at the bottom of the Builder panel. Select the chart, and scroll to the bottom of the Builder panel. Click on the Edit icon (L) to see the Predictive Forecast panel again. Review your settings, and make the required changes in this panel. For example, drag the slider toward the left to set the Forecast Periods value to 3 (M). Click on OK to save your settings.

The chart should now display the forecast for three months--January, February, and March 2016 (N):

Adding a time calculation

If you want to display values such as year-over-year sales trends or year-to-date totals in your chart, then you can utilize the time calculation feature of SAP Analytics Cloud. The time calculation feature provides you with several calculation options. In order to use this feature, your chart must contain a time dimension with the appropriate level of granularity. For example, if you want to see quarter-over-quarter results, the time dimension must include quarterly or even monthly results. The space constraint prevents us from going through all these options. However, we will utilize the year-over-year option to compare yearly results in this article to get an idea about this feature. Execute the following instructions to first create a bar chart that shows the sold quantities of the four product categories. Then, add a time calculation to the chart to reveal the year-over-year changes in quantity sold for each category.

• As usual, add a blank chart to the page using the chart option on the Insert toolbar.
• Select the Best Run model as Data Source for the chart.
• Select the Bar/Column chart from the Comparison category.
• In the Measures section, click on the Add Measure link, and select Quantity Sold.
• Click on the Add Dimension link in the Dimensions section, and select Product
• as the chart’s dimension, as shown here:

The chart appears on the page. At this stage, if you click on the More icon representing Quantity sold, you will see that the Add Time Calculation option (A) is grayed out. This is because time calculations require a time dimension to the chart, which we will add next.

• Click on the Add Dimension link in the Dimensions section, and select Date to add this time dimension to the chart. The chart transforms, as illustrated in the following screenshot:

To display the results in the chart at the year level, you need to apply a filter as follows:

• Click on the filter icon in the Date dimension, and select Filter by Member.
• In the Set Members for Date dialog box, expand the all node, and select 2014, 2015, and 2016, individually. Once again, the chart changes to reflect the application of filter, as illustrated in the following screenshot:

Now that a time dimension has been added to the chart, we can add a time calculation to it as follows:

• Click on the More icon in the Quantity sold measure. Select Add Time Calculation from the menu.
• Choose Year Over Year.

New bars (A) and a corresponding legend (B) will be added to the chart, which help you compare yearly results, as shown in the following screenshot:

To summarize, we provided hands-on exposure on predictive forecasting in SAP Analytics Cloud, where you learned about how to use a trend time series chart to view net revenue trends throughout the range of a year.

If you enjoyed this excerpt, check out the book Learning SAP Analytics Cloud, to get an understanding of SAP Analytics Cloud platform and how to create better BI solutions.