Skip to Content
Product Information
Author's profile photo Tharachandar B

Time Series Forecast Prediction Scenario in SAP Analytics Cloud


SAP Analytics for cloud is cloud SaaS based Business Intelligence tool provided by SAP. It was formerly known as BusinessObjects for cloud. It provides all the key functionalities of an Analytics tool to SAP business users. SAP Analytics cloud provides prediction as an inbuild  functionality.

For this example, we will be using a sales data source from Kaggle. For information Kaggle, a subsidiary of Google LLC, is an online community of data scientists and machine learning practitioners. Kaggle allows users to find and publish data sets, explore, and build models in a web-based data-science environment, work with other data scientists and machine learning engineers, and enter competitions to solve data science challenges.

This model will be useful for the organisations to predict their future sales data and plan for profit over the years. Source data we use here will the historical data for sales and with Time Series Forecast Prediction Scenario we are going to predict the future sales.

Sales Forecasting on SAC Classification:

The version of SAC used here is 2021.20. SAC is integrated with Machine Learning Algorithms by using this we can perform Smart Predictions. As mentioned before, in this model we going to predict and forecast the sales data.

Now, let’s start creating the model step by step.

Importing the data set

Go to the data set tab and click on the Create New from a CSV or Excel file and import the sales data.



Sales data set

Data set Overview

After importing the data, you can now be able to see the raw data in the middle and overview of the data in the right-side panel. The right-side panel display the total number of rows and columns we have with the raw data.


Data set Overview


Now switch to Columns tab to get more details about each column. The column view consists of Dimension Properties, Data Type, statistical type, data distribution and validation.



Statistical type

SAC has a functionality to automatically detect the data type and Statistical type. In some cases, SAC may fail to detect the data. So, it is always recommended to check the data type and statistical type and manually make changes if needed.

Statistical types and their purpose.

  1. Nominal : Unordered, discrete values (ex. Categories)
  2. Ordinal : Sortable, discrete values (ex. rankings)
  3. Continuous : Numerical, sortable, continuous values (ex. salaries, dates)
  4. Textual : Nominal values containing text (ex. sentences)

Once all details been checked Save the data set.

Create Predictive Scenario

Navigate to the Predictive Scenario tab and select Time Series Forecast since we want to forecast the sales data with respect to historical data.

For simple understanding, you want to forecast numerical values over a time period taking into account variables that may or may not be correlated.

Example: Forecast the volume of ice cream sold by a retailer for a future period using historical sales information, along with month and temperature data as variables that influence demand.


Time Series Forecast

Create a Predictive model with the sales data. Select Sales data in Time Series Data Source.

In Predictive Goal for Target select the field to be predicted based on Date. Specify the Number of Forecast Periods. For this example, let’s select 5.


Predictive Goal

In Predictive model training select Train Usage. Either we can choose All Observation or Window of Observation.

Train Usage is used to define the range of observations that will be used to train the predictive model. For example, you may want to ignore very old observations to avoid that your predictive model learns based on an obsolete behavior.

For this example, let’s select Window of Observation and provide Window Size as 5 and for Years and Until as Last Observation.


Predictive Model Training

Now click on Train and Forecast. The model will be trained and forecast the prediction.

After successful training of model forecast will be displayed.

MAPE % for the trained model is 0.87%. Any MAPE% <10 will be considered as Very Good.

MAPE % Interpretation
< 10% Very Good
10% – 20% Good
20% – 50% Ok
>50% Not Good




Forecast vs. Actual




This Blog shows step by step procedure to create a Time Series Forecast Prediction Scenario in SAP Analytics Cloud. I hope this blog help to understand the SAC Prediction with Time Series Forecast.

Thanks !

Assigned Tags

      1 Comment
      You must be Logged on to comment or reply to a post.
      Author's profile photo Antoine CHABERT
      Antoine CHABERT

      Hello Tharachandar B   many thanks for creating the blog!

      Some suggestions & comments from my side, some of them similar to what I mentioned here:

      • Kindly share the dataset and/or mention data source (Kaggle link) you used for this example. This will help readers recreate & leverage your example.
      • suggesting you can add the custom tag Smart Predict and the standard tag SAP Analytics Cloud, augmented analytics so that it's easier for SAP community readers to find your blog (more below)
      • A quick comment on "MAPE % for the trained model is 0.87%. Any MAPE% <10 will be considered as Very Good.". Actually in practice, the MAPE threshold depends on the exact prediction error that businesses would allow (or in other words the accuracy level they would expect compared to their existing processes). The Expected MAPE definition can be found here Quoting the help "The Expected MAPE (MAPE - mean absolute percentage error) is the evaluation of the error made when using the predictive model to estimate the future values of the target".

        Expected MAPE of 12% indicates that the error made when using a forecasted value will be of more or less 12%.

        An Expected MAPE of zero indicates a perfect predictive model.

      • Why not deepening the use case and and trying out our entity feature (segmented forecasting) - see an example here:

      Again, thanks for sharing your content with the community!

      Kind regards

      Antoine Chabert (SAC Smart Predict product manager)

      Here the 54 Smart Predict blogs to date

      Here the 73 blogs related to SAC, augmented analytics