Announcing the release of SAP Predictive Analytics 3.3
We are delighted to announce the release of SAP Predictive Analytics 3.3. This new release is available for download on SAP Software Center.
Python API for Automated Analytics
Automated Analytics provides a set of machine learning functions such as classification, regression, key influencers, clustering, forecasting, recommendation and link analysis to easily answer business questions with no need for extensive training or data science skills.
Python is an open-source programming language widely used within the data science community.
Through its Python API, SAP Predictive Analytics 3.3 allows developers to build powerful machine learning applications. They can now use Python as a scripting language to create, train, debrief and apply automated models.
The new API makes automated functions available not only within a Python script but also within a web-based interactive computational environment like the Jupyter Notebook.
Publish Predictive Models to SAP Business Applications
Predictive Factory is a web-based application that simplifies, accelerates and automates the predictive modeling process.
Predictive Analytics integrator (PAi) is a framework for applications built on SAP HANA to use and manage predictive models. It is a component of SAP Predictive Analytics.
With version 3.3, Predictive Factory users can seamlessly publish customized classification and regression models to a PAi server for integration into SAP business applications such as SAP Fraud Management or SAP S/4HANA Sourcing and Procurement.
Data Privacy Support
SAP Predictive Analytics 3.3 comes with security features and specific data-protection functions to support compliance with the relevant legal requirements and data privacy.
Predictive Factory delivers new capabilities to implement a data retention policy and limit the accumulation of unnecessary data storage. Administrators can schedule the automatic daily deletion of data. A new log file helps them trace data deletion activity. Administrators can also keep track of active user accounts and user access to forecast and signal analysis data.
Other Features
Before you go and download the 3.3 release let’s see other features it brings.
You can deploy Predictive Factory 3.3 on Linux systems as well as Windows.
User authentication is now consolidated with extended LDAP support in Predictive Factory.
The SQL scoring equation, for classification automated models, includes new prediction outputs like the decision and the contributions.
interesting to know, thanks for sharing!
Hello Marc, I am not able to get the connection between SAP PA and a model written in jupyter notebook. Apologies for ignorance. Please make me understand the following:
If I have a model created in jupyter notebook, how can I deploy it to SAP PA? Which module of SAP PA should I use and which component of SAP PA allow me copy the code from jupyter and paste in SAP PA before I train the model in SAP PA with access to HANA DB?
Regards.
Hi Debashish,
The easiest way to build, train and deploy predictive models on SAP HANA is to use either PA Desktop or the web application, Predictive Factory.
Jupyter is more suited for research and modeling experiments. You can use the notebook as a presentation or education tool.
In case you have not seen it, here is the webinar on SAP Predictive Analytics 3.3.
https://www.youtube.com/watch?v=Rt16MxuZa8U
The Python part starts at 2:30
Hello Marc,
Thanks for your response.
The video talks about Python scripting using "usual" environment. As data scientists create their models in text editors[like notepad++/atom etc], I understand from the video that Python scripts can be deployed to PA3.3 environment. In the first execution, you have executed the model[including "aalib"] via the command prompt. In the second execution, you have copied the same code[including "aalib"] into jupyter notebook. In none of the cases, you actually have entered PA3.3->automated analytics. Hence, the Python API was operational talking to PA3.3 Automated Analytics functionality.
With environments like jupyter, I have the option to ask for user dialogue to enter a URL to fetch data and analyze. I believe if the Python code contain input functionality[for URL], the same will be taken care of by PA3.3 with the help of "aalib". Please correct my statement if wrong.
Question: Do SAP offer Predictive Factory trial, as well? I was not able to find in web, if at all.
Regards.
Hi Debashish,
Aalib in its current form connects either to files (txt, csv, xls, SAS), either to databases via ODBC. It does not support URL.
For the Trial version including Predictive Factory, please follow the instructions from Thierry Brunet.
https://blogs.sap.com/2017/10/25/getting-started-with-trial-version-of-sap-predictive-analytics-professional-edition-3.2-with-a-cal-account-on-aws/
It is the PA 3.2 trial. The trial for PA 3.3 is under construction.
Best Regards,
Hello Marc,
Thanks for sharing..Very informative. We have time series forecasting models in PAI..
Thanks,
Hi Marc
thank you for your hard work. I have a question about SAP Predictive Analytics. Is it possible to publish a model version to a Pai ABAP Application, which I created with Expert Analytics? Cause I want to create a new predictive scenario with an expert analytics model and embed that within my S/4HANA application.
I would be very grateful for your reply.
Thanks,
Lee
I am not aware of such functionality “Publish Predictive Model to PAi” in SAP Expert Analytics. I think, when using Automated nodes in the pipeline, SAP Expert Analytics provides an export in SAP HANA as stored procedure. Note that SAP Expert Analytics is in maintenance mode, as per https://blogs.sap.com/2020/01/27/sap-predictive-analytics-maintenance-policy/
One alternative would be to use a SQL script to train the predictive model:
and then apply the trained model via ABAP-Managed Database Procedures (AMDP):
APL documentation is here: https://help.sap.com/viewer/p/apl
Another alternative would be to use SAP Data Intelligence https://www.sap.com/products/data-intelligence.html
The new name for PAi is ISLM: https://blogs.sap.com/2020/07/22/building-into-the-digital-core/