Predictive Modeling in SAP Analytics Cloud #ASUG @...

Technology Blog Posts by Members

Explore a vibrant mix of technical expertise, industry insights, and tech buzz in member blogs covering SAP products, technology, and events. Get in the mix!

3,214

This was an ASUG webcast today. As usual, the SAP legal disclaimer applies - anything in the future is subject to change. Slides are posted here. Register for the replay here

Note: SAC abbreviations mean "SAP Analytics Cloud"

Abstract:

See how predictive in SAP Analytics Cloud delivers an integrated solution capable of solving real world use cases, which brings the best of BI, Planning and Predictive in a single platform. In this webinar we will show how SAP Analytics Cloud customers can build predictive models, consume the results of those models in BI and Planning scenarios and put the models into production in Business Applications. We will also cover our vision for predictive within Analytics Cloud.

Source: SAP

"SAP Predictive Analytics enables customers to create and manage large scale Machine Learning models against SAP application data and to embed the predictive outputs and insights where they drive the most business value: directly in the applications where business decision makers are working." - Source: SAP

The above shows the current state with on premise Predictive Analytics solutions.

Source: SAP

From Q1 2018 all net new Predictive Analytics development is being redirected into SAP Analytics Cloud. SAP is building a simple and compelling user experience for predictive workflows in SAP Analytics Cloud that will compliment and enhance existing Business Intelligence and Planning capabilities. The focus is to provide simple self service Data Science capabilities including support for data ingestion, model creation and model management.

The target user of this development will be the SAP Analytics Cloud user. SAP's goal is to provide professional data scientist quality predictive models to Data Analysts and BI user through a simple interface that enables them to augment their planning and BI workflows. It will be targeted at the citizen data scientist, business analyst - not the data scientist. It will be provided in a consumer level experience (source: SAP)

Automated data science is the way going forward, by 2019 citizen data scientist will surpass data scientist in output, as projected per Gartner

It will support large scale machine learning and support embedding models in business applications

Source: SAP

Move to Analytics Cloud is to provide a single application. Typically predictive solutions require a lot of set-up; Analytics Cloud provides multiple sources; Predictive will support those sources. There is no need for IT setup.

SAP's long term vision is to provide a tool with all of the functionality of SAP Predictive Analytics with 100% in the cloud. The plan is to "Enable a Data Scientist or data analyst to work 100% in a SaaS solution." (source: SAP)

Part of SAP Analytics Cloud includes the following:

Data connectivity

Data ingestion

Model authoring

Model management (source: SAP)

Source: SAP

Future plans are subject to change.

Initial release is planned for June around SAPPHIRENOW

They will provide predictive modeling, classifications (answer which customers will buy a product or churn), will support regression (when deal is complete), time series forecasting - segmented forecasting (sales over time, build by segments)

Support CSV/XLX formats, S/4HANA

Consume predictive scores in Analytics Cloud

Q3 - the plan (subject to change) Extend support for predictive modeling, more complex models, more than 1 technique, ability to export predictive results

Extend application integration - planned around August around S/4HANA release

Q4 - plan to introduce hybrid support - live connection to data - on premise data supported so data is not transferred to cloud and score data in source system

Next year the plan is to include data preparation. The rest of the session included demo which you can watch in the replay here

Question and Answer

Question	Response
Can we use CDS Views (Consumption Views) as sources to the SAP Analytical cloud and do predictive modeling?.	We will support connections to HANA (not in inital release) - probably Q3. S4 connection will support CDS views
Do you have a demo on Predictive Modelling?	happening soon here, live!
Is there any plan to add feature generation to SAC?	Yes, plan to add the ability to build analytical record used to train a model
ABAP CDS voews	S4HANA connectivity will support CDS views
Can we connect external ML tools like R	Initally not support extensibility
what does this mean for Predictive Factory	Predictive Factory continues to be available/ supported - future development is focused on cloud
Does it have everything that is available in PAL but in cloud?	focusing on answer business questions; rather than provide a catalog of technique - focus on business question rather than a data scientist
What data sources are planned for Predictive analytics?	plan to support all SAC and data sources introduced for SAC; for live - support HANA, Hadoop
What about unstructured data from platforms like hadoop? will it be possible to consume models from SAP Vora?	planned support for hadoop
Does it support Unstructured data?	Typically algorithms are suited for structured data
maybe already asked, but showed up late. Is functionality you are showing in SAC, or planned to be there	Planned functionality for around the June/SAPPHIRENOW timeframe - thank you for joining us today
In future, do you have plans to include the social media( like Twitter, Facebook), to do any predictive analytics?. Fetching these data to Predictive models is the key though but want to understand the view from SAP on using these social media as sources?.	no answer to; depends on how SAC integrates with those data sources
What are the optimization techniques in SAC? seems your screen choking when processing excel file data	Predictive today builds multiple techniques - build iterations of model for optimization
Will it be possible in future to consume unstructured data from hadoop via SAP Vora?	On roadmap to connect Hadoop - no details on that
Is your data for this demo coming from C4C?	He said it is an Excel file
Where is excel populated from?	this is a sample data set
can you have demo on the abap cds connection to choose the table	not in this system, not available yet; target for SAPPHIRENOW
Is there any plan to add automated feature generation?	yes, plan to do this
and how do you put the entry back into the same table	not possible to do this directly
Can I say this is HANA PAL in Cloud?	no, not provide full PAL coverage but to answer business questions
Will you support live unv data as a data source for Predictive analytics	Yes, planned for early next year with a live connection
Can you use TensorFlow iwth this?	use algorithm for business question rather than data scientist techniques
Is CPI/HCI used for SAP Predictive Analytics Integration?	not really - use same connectivity as SAC
If you are focussing on business questions, are you planing industry specific release ? Like specific scenarios for Resources industry ?	looking at industry content in SAC and plan to incorporate predictive into that
So this is aimed to Business Audience?	aimed at citizen data scientist
how is this integrated with SAC planning functionality	planning integration not demoed here but something to be introduced quickly