Earlier this month SAP’s Scott Leaver provided an ASUG webcast on Predictive Analysis. My notes are below. As usual, things listed below are subject to change as the product is not yet generally available.
Figure 1: Source: SAP
Predictive is an important requirement to help predict and “act before react”. The top box shows they are delivering to market a “new perspective” for analysis
Scott said it provides a “modern UI to do analysis and visualizations”
Figure 2: Source: SAP
Figure 2 explains what predictive analysis is.
Scott also reviewed some common myths that you need a data warehouse in place for predictive analysis (not true, he says) or that you need a PhD to do it.
Figure 3: Source: SAP
Figure 3 shows on the right a list of algorithms. You can drag and drop on canvas. Scott said that R open source library is no cost as it is a community approach.
Predictive Analysis ships with its own R algorithms – see “R” indicator on the right side. No scripting or coding is required.
Figure 4: Source: SAP
As Figure 4 shows, after modeling, Predictive Analysis has a catalog of visualizations. It uses CVOM (charting visualization object model) for visualizations
It is a desktop resident tool – runs locally on desktop on a Windows 7 desktop. It is an eclipse-based tool.
It takes a variety of approaches and share them, using PMML (Predictive Modeling Markup Language)
Figure 5: Source: SAP
Process driven approach starts at “12 o’clock” (in Figure 5) by identifying data, then preparation (“standard data services” and then step 3, data processing. The last step is the visualization, the “so what” to build a campaign.
Figure 6: Source: SAP
The tool Predictive Analysis works on the Windows 7 Desktop
On the left is running it on the Desktop and then pull in from various systems
It writes the model back to datastore
On the right is the “non-Hana approach”, running it locally, via CSV or relational data source, and do some analysis and write back to modeling repository.
It does not require Hana to run but it is better for large data volumes.
Figure 7: Source: SAP
Algorithms are provided with samples / algorithms. If you wanted to do clustering you would use K-means
This is a learning mind map to use.
Figure 8: Source: SAP
The tool has a catalog of visualizations. You don’t have to use another client
If you want a more ad-hoc approach with a BI application and Figure 8 above shows what you can use
To join ramp-up, go to http://bit.ly/PARAMPUP
Subset of Question & Answer:
Q: What is the twitter hashtag for Predictive Analysis?
A: Scott says there is – he’ll create one #SAPPredictive
Q: Does it have BICS connectivity for BW?
A: Not yet; if you want BW, use JDBC to connect to BW or pull BW data into HANA
Q: Will Predictive Analysis only be released for Windows? Are there plans for a version for the Mac?
A: In the roadmap; haven’t defined when
Q: “Write back the results of analysis to SAP applications” Can you provide more details please…how and which applications are supported?
A: High level you have data writer option – use standard .CSV and write back to a set of tables and use a client application to read that data
Q: #SAP Predictive Analysis Algorithms: Does the tool have inbuilt help to guide users regarding which to use when?
A: try this link http://help.sap.com/businessobject/product_guides/SBOpa10/en/pa_user_en.pdf