Note: this was posted on the SCN Beta space which is going way – so I am posting this here. This was an SAP webcast last month
SAP will continue to invest
Help deliver predictive at scale, with accurate models
Predictive factory components (replace Model Manager)
Deploy into applications so generate ROI on an ongoing basis
Successor to Model Manager
Could be use for data prep
Key benefit is to manage thousands of predictive models
Platform allows for collaboration
Big data is Hadoop and Spark
Release 2.5 support model training on Spark in Hadoop
Delegated machine learning to this
Release 3.0 support regression models on Spark, and user structures
Automated Analytics & APL
Train and apply using attribute views/master data
Export as a new model chain is new
Question and Answer
Q: when will PA 3.0 be GA?
A: Hello, we expect this release to happen in the second quarter of 2016 (already GA now)
Q: is that a new server software or does it require hana?
A: Hi, Predictive Factory is currently a on premise server solution based on Automated
It can leverage Hana capabilities, including delegation of the model training to Hana and Hana Views support but it works also with other database types (through ODBC) and with file based datasets
Q: Does PA Factory need to be installed on a specific server (runs on Windows and Linux)?
A: Predictive Factory has its own setup and relies for its scheduling capabilities on a simplified version of our BI platform. Predictive Factory depends on an Automated Analytics server.
It can be installed only on Windows for the time being.
Q: Is this new release will have a new and unified UI?
A: The Predictive Factory will have new UI.
Open question as to whether this would be available on the CAL
Link to roadmap: https://websmp205.sap-ag.de/~sapidb/011000358700001160102012E.pdf
Upcoming ASUG Predictive webcasts: