Predictive Analytics 2.0 – What is New - ASUG Webcast

[TammyPowlas]

This was an ASUG webcast given by SAP yesterday with charles.gadalla and abdel.dadouche

Figure 1: Source: SAP

Bring together two code lines, blue bubble, InfiniteInsight (blue) and SAP Predictive Analysis will be combined into one bubble on the right as the expert and automated come together

Figure 2: Source: SAP

Latest releases become first release in 2.0 – Predictive Analytics 2.0 is the legal successor

It is not bound to the Lumira release cycle

SAP is decoupling code; it will not release on monthly cycle that Lumira

Figure 3: Source: SAP

There are 4 installation packages shown together

Desktop version automated analytics (workstation from InfiniteInsight and expert from Predictive)

Client drives server component (data manager, automated algorithms, scoring, social recommendations)

Expert Analytics only available in desktop installer

Predictive Analysis is on workstation side; there is no server component for Predictive Analysis

Figure 4: Source: SAP

Figure 4 shows the installer scenarios

Please note you need to uninstall old PA to install PA 2.0 as the unified installer will overwrite the libraries

Lumira will need to be uninstalled

Going forward the products will co-exist; for the desktop this includes workstation, client, server, factor

Figure 5: Source: SAP

Figure 5 shows the unified Launchpad

The automated analytics comes from InfiniteInsight

For classic Predictive Analysis  go to Expert Analytics

Figure 6: Source: SAP

Please note that the 30 day trial for Predictive Analytics 2.0 is not available yet; it is coming soon.

Figure 6 shows the keycodes; SAP will work with you on this; SAP has unified the license keys

Figure 7: Source: SAP

Where there were empty segments is now taken care of but preserve model integrity

On the bottom of Figure 7 they have smoothed out the curves without the breaks and segmentation

Figure 8: Source: SAP

Figure 8 shows Hadoop and Spark support; the SIMBA driver is used

Date support is shown in Figure 8; limitations are due to Spark

Figure 9: Source: SAP

Figure 9 covers if you are on the HANA "side of the fence" and what is new with the automated predictive library

The libraries from KXEN are exposed to HANA as APL - more to come

They will create, train, and apply models in HANA

Applications can invoke these

This is giving HANA the flexibility to use

Figure 10: Source: SAP

"Everybody loves BW"

SAP Predictive can do this in the offline mode

It can use the BEx queries as shown in Figure 10

Figure 11: Source: SAP

Figure 11 shows what is new in the PAL components

You can write your own custom PAL components

Similar to the way you could create your own custom R component

R library is the open source library for free

The PAL was created by HANA team to invoke similar R algorithms but you can call it in memory; you get a huge execution

You can prepare your own custom PAL algorithm using the SAP Predictive Analytics 2.0

Question and Answer

Q:  Will the Lumira rooms such as Compose/Share be removed from the "Expert Analytics"  portion of the product?   It seems that the 2 products are meant to be more separate than before since the SAP PA will not be on the Lumira dev schedule

A:  No; as analysts, need to share; step 1 unify code bases - unify code line, deploy/share execute - move to UI5 interface coming.

________________________________________________________________

Q:  can you elaborate on the ease of use for marketing etc. business users without much background knowledge?

A:  In automated modeler, so you don't have to choose algorithm. Business Analyst/marketing doesn't have to know model details

________________________________________________________________

Q:  Users of Lumira server/cloud will not be able to use SAP PA 2,  so that is why I am wondering if the compose/share part will be removed.  Lumira must be on same level as cloud and server.

A:  You can publish to both - same code line; in terms of Predictive having a server - not available today, using KXEN.

________________________________________________________________

Q:  You referenced a blog.  Can I have the link?

A:  http://scn.sap.com/docs/DOC-62547  

________________________________________________________________

Q:  Lumira and SAP PA will be at different versions  so SAP PA will not be at the same version level as server or cloud.

A:  Yes - this is true, so you need to be aware of the notes/release restrictions

________________________________________________________________

Q:  If all the processing all within HANA, is there a way to prevent too much resources being used on HANA?

A:  HANA Studio has ways of identifying resources required; Predictive is passing back an algorithm

________________________________________________________________

Q:  do we provide help to interpret the automated modeler results and suggest next steps etc. for te hbusiness user(again :-)...I think that is most powerful feature of PA 2.0 allowing business/marketing folks to utilize predictive etc.

A:  In automated mode, send the output - demoing now

________________________________________________________________

Q:  Can't find the Trail Version of 2.0 on SAP Store

A: That is coming soon - not available yet

________________________________________________________________

Q:  The product (SAP PA and SAP Lumira are only supported at the latest version.  If the Lumira Cloud is at version 1.23,  then the desktop version needs to be at 1.23  There is where my issue with SAP PA not being on the same maint schedule comes from.

A:  True - decoupling from Lumira, on a slower release cycle; that does not mean they are decoupling from the capability. Will start to OEM the solution so you are not disadvantaged.

________________________________________________________________

I asked for a sneak preview of what is coming to ASUG Annual Conference/SAPPHIRENOW in May.  SAP Predictive Analytics 2.2, something with the HANA Cloud Platform and some "undercover items"

If you are an ASUG member you can watch the recording and demo at https://www.asug.com/discussions/docs/DOC-40642