Skip to Content
Product Information
Author's profile photo Dalibor Knis

Machine Learning Cockpit for SAP Business ByDesign

Do you think there is a use for Machine Learning in Small and Medium Enterprises?

Would you like to be able to design predictive scenarios on your own directly in SAP Business ByDesign? At no extra cost, no external service to use, no developer or data scientist required?

If you have this ambitions then this blog is for you.

 

In SAP Business ByDesign release 2211 we are introducing Machine Learning Cockip.

 

Overview

Machine Learning Cockpit for SAP Business ByDesign is SAP implementation of standard machine learning workflow within SAP Business ByDesign user interface.

Machine Learning Cockpit (MLC) brings Predictive Analytics capabilities to Business ByDesign by utilizing the power of Predictive Analytical Library (PAL) of SAP HANA of Business ByDesign tenant.

The tool allows users to create predictive scenarios by using data of Business Analytics data sources, on which PAL machine learning algorithms try to learn the model in the training process. This model can be applied to new data to make actual predictions. The predictions can be then consumed directly in the Business ByDesign business documents, in Business Analytics, or in PDI solutions (through provided API).

Relevance

Machine Learning Cockpit (MLC) is relevant to partners who want to provide their customers intelligent solutions that are based on analytical predictions.

Benefits

Machine Learning Cockpit (MLC) implements standard ML workflow directly in Business ByDesign user interface. Users are guided through the ML workflow in an easy-to-use interface. Thus, users can focus on the business context of their predictive scenarios and do not have to deal with the complexity of Predictive Analytical Library (PAL).

The MLC is a fully embedded solution. All the data used in the pre-processing, model training and predictions never leave Business ByDesign tenant.

 

Stay tuned. More blogs, videos and tutorial to be followed.

 

 

 

Assigned Tags

      5 Comments
      You must be Logged on to comment or reply to a post.
      Author's profile photo Cesar Cano
      Cesar Cano

      Hello dalibor , I hope you are well.

      I understand that there are 3 ways to consume the prediction results, through the use of the data source, use of the integrated component and through the use in extension fields, how could we use it concretely in a field? I have made tests however I can not select any field created in the prediction execution, is there any other step or additional process to consider to be able to reflect the result in a field?

      Regards

      Author's profile photo Dalibor Knis
      Dalibor Knis
      Blog Post Author

      Hi Cesar,

      As general rule, extension fields must be created in the same business object/ business context as is the datasource anchor. If you did it this way then you have maybe stumbled upon some technical issue which we would love to resolve asap.

      Can you be please share details regarding what dat source you used and where exactly did you create extension fields?

      kind regards
      Dalibor

       

      Author's profile photo Aruna Thakar
      Aruna Thakar

      Hi,

      We would like to use this new functionality for our production planning process and Sales delivery process...We would like to calculate total processing time for production task completion, just for example. Or send email notification to customer proactively for delayed shipment of any sales delivery??

      Is it possible to use this feature for such scenarios?

      TIA

      Author's profile photo Dalibor Knis
      Dalibor Knis
      Blog Post Author

      Hi TIA,

      from your description it seems that prediction of Total Processing Time is similar to prediction of Customer Invoice Payment Delay and Purchase Order Delivery delay.

      Please note, all these use-cases are primarily Regression problem, rather than Classification problem. It means that you will need to cluster values of Total Processing Time to distinct categories e.g. Short, Medium, Long, Extra Long and use this new calculated field as Target Field in MLC Scenario definition. See the above mentioned use cases as examples how to do it.

      However, you still need to identify suitable analytical data source(s) containing:

      1. enough data
      2. enough fields, containing data potentially useful for predictions
      3. fields from which the categorical Target Field can be calculated

      If you have all this then yes, it should be possible to use MLC for proposed use case.

      kind regards

      Dalibor

       

      Author's profile photo Aruna Thakar
      Aruna Thakar

      Yeah, I have all the data for the required calculations.....will check the process and let you know my findings.

      Thanks In Advance 🙂

       

      Regards,

      Aruna Thakar