#askSAP Analytics Innovations Community Call: Reimagine Predictive Analytics for the Digital Enterprise Webcast Recap Part 1
This was an SAP webcast held last week
The usual legal disclaimer applies
Source: SAP
Source: SAP
SAP’s Ashish Morzaria started with Predictive Use cases, how the digital economy has changed
The internet has moved from a push to a peer-to-peer network
Enterprises have invested to connect suppliers, systems, every transaction is recorded and digitized.
This information is shared with people in the supply chain.
Source: SAP
New economy encourages enterprises to turn digitize assets to their advantage
Inject predictive processes into help improve processes and decisions
How create new products, services and models? Use algorithms to analyze data that you have to develop a competitive edge.
Source: SAP
Companies embracing these technologies are winning
Companies who have invested in algorithms have more revenue, more profitable, more competitive
Source: SAP
Predictive Analytics accelerates process using automated techniques
Faster, without a single line of code, and models can be reused by analysts in organization
You can apply a segmentation to them to create derivatives of models
Workflows are repeatable, with guided algorithms
The data scientists can use guided workflow and do additional configuration and parameter tuning to understand results
Supports HANA and Hadoop and can push calculations down to HANA and Hadoop to reduce data transfer required
Source: SAP
HANA can run R scripts, an open source library
The issue with R is that it runs outside of HANA and this is inefficient
To solve, SAP created PAL, and PAL follows 80/20 – you can create algorithm using R, but common ones are run natively in HANA to enable in-memory execution
APL (automated predictive library) for relational sources, now in HANA
HANA can use native algorithms, APL, and automated capabilities to run
Predictive Analytics use all of this with one interface
Source: SAP
On the left you have traditional analytics
Bring into warehouse, analytical source, each step has a data extraction
Using HANA, with Predictive Analytics, federate queries/calculations to HANA layer
Flexibility of using R algorithm, mix and match between APL, PAL, R
Source: SAP
Smart data streaming includes the capability to apply scoring algorithm to streams, apply scoring “on the fly”
Source: SAP
Traditional data solutions start sampling, reducing efficiency and accuracy
When using Hadoop with a single source, limited by system with analytic engine
Hadoop + Spark allows doing things across a large number of nodes
Native spark modeling federates calculation to the model itself; have Predictive Analytics software orchestrate in Spark
Source: SAP
This is a deeper diagram showing how it works
The Spark engine is running underneath
Traditional analytics was using code
New predictive analytics has a wizard approach; workflow is similar to using a relational source
Source: SAP
Tool that works for both the data scientist and citizen data scientist
Source: SAP:
Predictive analytics is part of an analytical environment
Traditional analytics answers the questions on the left; very descriptive types of questions answering
What promotions should I run (on the right)
With predictive, understand pattern of past to apply to the future
Source: SAP:
Predictive analytics is part of an analytical environment
Traditional analytics answers the questions on the left; very descriptive types of questions answering
What promotions should I run (on the right)
With predictive, understand pattern of past to apply to the future
Source: SAP
Embedding in a workflow
Start with Predictive, create model, export model, and apply directly to a database (HANA, Oracle, etc.)
Once exported, any application using SQL will get a recordset out
Now any tool that can access database can report on it.
Every BI user can access score in this example
Source: SAP
Batch scoring; creating a retention campaign, address customers who are likely to leave
Take data that represents customers
Apply predictive model, take the score
Your BI user filters on those likely to churn
Questions:
Q: Are there any plans to make R scripts first class citizens by embedding them in HANA?
A: Plans to bring them to HANA, due to legal issues, will run on HANA but still be a separate process
No plans to make it an in-memory engine inside HANA itself
Q: BusinessObjects Predictive Analytics a separate license?
A: Separate product, licenses are included in BusinessObjects Enterprise Premium announced at SAPPHIRENOW
Capabilities exist in BusinessObjects Cloud for Predictive – more of an embedded case
Q: How Graphx – plan to embed in HANA?
A: Able to do link analysis in HANA using tools
Poll: Source: SAP