Approaches and the flow – leveraging Machine Learning and Predictive Analytics for SAP S/4HANA (Updated Feb 4th 2023)
Part 3 of the blog series:
A recent podcast conversation with Hadi Hares from SAP is here.
Another podcast conversation with SAP Experts Priti Dhingra and Abhishek Mishra on ISLM enhancements with use in SAP S/4HANA is here.
Another podcast conversation with SAP Experts Robert McGrath and Antoine Chabert here explains about the Smart features with SAP Analytics Cloud.
A Podcast on the architecture concepts and approaches is here.
In this blog, I will be focusing on the different approaches available to leverage the SAP S/4HANA functionality with Machine Learning and Predictive Analytics algorithms. This is a continuation to my earlier series of blogs on the topic – in the last blog we focused on the architecture and did a bit of a deep dive into the same.
Reviewing a bit on the different approaches while leveraging the SAP S/4HANA functionality, here is an updated view point on the different approaches.
In the above picture, you will notice the different approaches of leveraging ML and Predictive Analytics with SAP S/4HANA. In the context of all the above approaches, we shall have detailed discussions in our upcoming blogs. We have briefly discussed about the first 3 approaches of building into the digital core, expanding around the digital core and extending the digital core. The 4th approach is basically explaining how you can extend some of the out-of-the-box released predictive models. Stay tuned for more updates in this context of the various approaches available to leverage the different ML and Predictive algorithms around SAP S/4HANA.
Now let us deep dive into the process flow involved around these different approaches.
Embedded Predictive algorithms in SAP S/4HANA:
A brief walk thru’ into the process flow while embedding Predictive Algorithms into the SAP S/4HANA business processes.
As part of the Architecture blog, we have seen how the predictive models in SAP S/4HANA are exposed thru’ the SQL Script procedures and made available with the CDS views to be embedded in SAP S/4HANA business processes. Now as shown in the above picture, let us look into the various steps involved in the embedding of these predictive models in to the SAP S/4HANA business processes. On the left-hand side you will notice the various data sources that can be leveraged to procure the data.
In the first step, data preparation happens with the Data manager, procuring the data from one of the data sources that is available. Secondly, the predictive model is built leveraging the ISLM framework using the automated model or expert model or even external models. Later this model is deployed into the S/4HANA business processes using the ISLM framework to utilize the model management for creating additional versions of these models. Finally leveraging the ISLM (Intelligent Scenario Lifecycle Management) framework, the Intelligent scenario management model is created for this model. The corresponding CDS view for this Intelligent scenario is then embedded into the application logic of the ABAP objects in SAP S/4HANA. You can learn more about creating an Intelligent Scenario from the SAP Help Portal. You will also learn more about creating an Intelligent Scenario, Intelligent Scenario management model etc., in the context of an example use case in my upcoming blogs and will be referenced here.
The data input and data output parameters are available as part of the intelligent scenario created and various different intelligent scenario management models can be created with different datasets and tested out for the results since this is already embedded into the S/4 business process. We will discuss about the functioning of these models in the context of the different use cases in a later blog. One important point to note is that you will need to use the SAP Fiori Apps with the Analytics specialist role to train and apply these embedded predictive models. Though this will be discussed at length in my upcoming blogs, here is a quick reference to the SAP help portal on how to configure the consumption of these intelligent scenario management models.
Consuming ML services on SAP Business Technology Platform:
Now let us look into the process flow of how ML services are consumed from the SAP Business Technology Platform by the SAP S/4HANA business processes in the context of leveraging SAP Data Intelligence and/or SAP AI business services. Prior to leveraging SAP Data Intelligence, the legacy Leonardo ML foundation framework (now transitioned into SAP AI Core and SAP AI Launchpad) is utilized to build the training models, deploy them and create the inferences which would help configure the communication scenarios with the SAP S/4HANA business processes. This helps the deployed ML services on SAP BTP to be consumed accordingly by the SAP S/4HANA business processes.
As of the latest update in Q4 2022, SAP strategy is 3-fold:
- Use the pre-delivered and pre-configured Intelligent Enterprise solutions explained as the out-of-box scope items whether embedded or side-by-side models.
- Use the AI Business Services that are available as ML services built on SAP BTP and configured to run with some of the SAP applications like SAP S/4HANA, SAP SuccessFactors etc. The SAP AI Business Services use SAP AI Core and SAP AI Launchpad to build the ML models, monitor and productize with SAP Applications.
- Finally, the 3rd strategy is to directly use the SAP AI Core and SAP AI Launchpad that are now available as GA in the Q4 of 2022. The customers and partners can use these SAP AI tools to build their own models or bring-in external models. But it is upto the customers and partners to maintain these AI models built with the tools SAP AI Core and SAP AI Launchpad.
So, as explained in the above figure, SAP AI tools like SAP AI Core and SAP AI Launchpad can be used in building the ML models, deploying and monitoring them. Customizing AI models, continuous delivery and consumption of the AI services are the key aspects here.
A few things to note while leveraging the SAP AI Core and SAP AI Launchpad:
- Accelerated Performance via GPUs if the demand is there
- Board range of resource Plans, so you can use what is in need for your AI use cases
- Run complex containerized (parallel) workloads
- Autoscaling gives you full flexibility to define how much performance is in need.
We will discuss in more detail in the blog related to “Enhancing the digital core with ML services” about the updates related to SAP AI Core and SAP AI Launchpad in the context of SAP AI Business Services along with the SAP DataIntelligence.
Let us now look into the new concept of leveraging SAP Data Intelligence for the creation of training and inference pipelines as provided in the picture below.
Please note: SAP DataIntelligence is predominantly going to be used for data management though it has the ML features that could be leveraged to build ML models.
As explained here in the above picture, there are 2 steps in the creation of the training process flow and inference process flow with SAP Data intelligence using the ML foundation services.
In the training process flow, the training data is saved and trained with the algorithms while saving the training model. Later in the inference pipeline, the ML services are ready for consumption by the SAP S/4HANA business processes – the trained model is read and applied accordingly. More details on the complete process can be understood in our detailed SAP help portal on how to leverage the Data Intelligence platform for creating and running the ML services. At a later date, we shall also explain the process flow in the context of leveraging the AI business services with the AI foundation layer and how they are consumed in SAP S/4HANA.
Leveraging Predictive services in the SAP Analytics Cloud:
Let us now look into the process flow involved in leveraging the predictive services in the SAP Analytics Cloud. This is the 3rd approach of leveraging the ML and Predictive Services using SAP tools for the benefits of the customers. As discussed in detail during the architecture blog, in this approach, some of the core algorithms available from the APL library can be leveraged in creating quick predictive analysis. Following are some of the functions that could be leveraged.
As you can see in the above picture, choosing the right smart feature in SAP Analytics Cloud is the key to perform the required functionality. Depending on the requirement and functionality you could do data acquisition or live connection with the SAP S/4HANA CDS views and do the analysis with the different smart predictive features. Stay tuned for more updates on this approach since we have a few (Proof-of-Concepts)PoCs going on with this approach!
A quick overview of the available smart features in SAP Analytics Cloud:
Smart Assist – Search to Insight – What is happening in my business?
Smart Assist – Smart Insights – What is behind this number?
Smart Assist – Smart Discovery – Why did this happen?
Smart Predict – Time Series Forecasting – What will happen?
Smart Predict – Smart Grouping – Which data points are similar?
Smart Predict – R Visualizations.
In our next blog we will continue our blog series with focus on the functionality and the scoping of how the Machine Learning and Predictive Services are applied into the SAP S/4HANA business processes.
Here are some quick links to the blogs in this series to give you a complete understanding of how Predictive Intelligence is infused into SAP S/4HANA.
- Resources and journey to machine learning with SAP S/4HANA
- Part 1 – Leveraging Predictive Intelligence with SAP S/4HANA
- Part 2 – Architecture and deep-dive of the different approaches around Predictive Intelligence
- Part 3 – Process flow leveraging Machine Learning and Predictive Analytics (this blog)
- Part 4 – Scope and functionality in the context of an end-to-end process leveraging ML
- Part 5 – Activating machine learning functionality for SAP S/4HANA
- Part 6 – Building ML into the digital core of SAP S/4HANA (Embedded ML)
- Part 7 – Enhancing the digital core with ML Services (Side-by-Side ML)
- Part 8 – Extending the digital core by leveraging ML with SAP Analytics Cloud
- Part 9 – ML Extensions to SAP S/4HANA processes
- Blog series – ISLM for machine learning with S/4
- Introducing the book – Implementing Machine Learning with SAP S/4HANA
Happy predicting the future!!
The simple option would be to connect Apache Spark via Hana JDBC to S/4. Wouldn't that be an option to discuss as well?
Yes and I shall discuss in my continuing blog series - that option would be part of the option1b and 4a, to connect with the core HANA PAL libraries. I have also discussed briefly in my previous blog on architecture which was written sometime back.
Thanks for the blog, this is really helpful!