Author: admin

SAP would like your thoughts on the name of a new SAP Support service we will be piloting soon.

If you would like to join a 30-minute virtual panel discussion on the topic, please email supportusability@sap.com with your name, location, and timezone. http://bit.ly/2i9LQwe #SAP #SAPCloud #AI

Source: SAP

You can watch the replay here

I also attended the same session live (but with a different customer)

Source: SAP

Forward thinking statements are subject to change

Source: SAP

Bring self service to business user and more sophisticated dashboard design

You get 3 applications with SAP Lumira:

Discovery Client – catering for the self service business use cases

Designer – more of a developer tool, enterprise ready applications with JavaScripting and a command engine in the product

Add-on to the BI Platform that brings together both on the BI Platform

There is also interoperability brought in, taking a self-service story to a development environment

Source: SAP

The two clients are sharing same graphical engine, output format (.lumx),

The graphic above shows the key components, including an in-memory engine (also in the desktop tools)

Reusing runtime for Design Studio for the HTML5

Source: SAP

Legal successor for products

“Seamless upgrade” for the tools

License is covered above

Source: SAP

Designer has a “rich roadmap” shown above

Left is the 2.0 release

December will ship for 2.1 (planned)

Source: SAP

This slide was not covered in the lecture, but I thought it I would include it

Source: SAP

2.0 has a Fiori style sheet; fits to the rest of the SAP styles

Source: SAP

Composites are a key feature for Designer; allowing designers to create their own composites, complex advanced components and reused across different applications

Source: SAP

Updates to the generic templates; get started quickly using the templates

Source: SAP

You can export to Analysis Office; you can keep navigating

Source: SAP

This was a quick outlook – ship scheduling/broadcasting capabilities; this is one of the gaps

Source: SAP

Commentary solution – comment on the dashboard layer and on the data; stored on a central place on the BI platform

Source: SAP

Three important themes for Lumira 2.1 are shown above

Close gaps with BEx Web

Source: SAP

Copy components

Source: SAP

Improved PDF export is shown above

Source: SAP

Close the 1.x gaps

Source: SAP

At run time you can hide/show components

Source: SAP

Linked Analysis – filter value and context between stories

Source: SAP

Filter on measure within a visualization and set context is planned for 2.1

Source: SAP

BW live analysis will be web-based in 2.1; the end to end live analysis to remove dependency on desktop for the end user

Source: SAP

consume Lumira on Fiori-fied BI Launchpad

Source: SAP

Governed self service for end users – a few more rights to export stories

Source: SAP

Planned for 2018 – 3 themes are shown above

Tighter integration with SAP Analytics

Source: SAP

Simplify workflows is part of the themes

Upcoming ASUG webcasts:

Oct 19 An exciting new future for SAP BusinessObjects Web Intelligence

Oct 26 Overall SAP Analytics strategy and direction 

Oct 31 What’s new in SAP BI Mobile and Roadmap

Nov 7 Cross-Solution Analytics with SAP Analytics Cloud

Nov 9 What’s New in SAP Lumira 2.0

Nov 20 What’s new in Analysis Office 2.6?

Nov 28 – Introduction to SAP Data Hub 

Nov 30 Refresh Expert Streams on Predictive Factory

Dec 5 – Update on Conversion Tools for SAP BW/4HANA

Dec 12 Best Practices for Automated Analytic http://bit.ly/2i7HTIv #SAP #SAPCloud #AI

This is the second blog in a series of three about enriching SAP BW with SAP Predictive Analytics:

* Introduction
* Technical Integration (this page)
* Steps of a Project

Technical Integration

In the previous blog we explained how SAP Predictive Analytics adds value to your SAP BW. In this document we look at the technical side of things.

There are two different options available on creating forecasts.

* They can be created in real-time.
* Or the forecasts can be calculated on regular intervals through automated scheduling.

This blogs describes the scheduling concept. Forecasts are produced on a regular basis and written into SAP HANA. This article is focusing on the technical workflow, assuming for example that you have already identified a clear business case for a specific prediction.

 

Technical prerequisites for this scenario are:

* SAP BW must run on SAP HANA. You will see that data is accessed through Calculation Views in SAP HANA. If your SAP BW is running on another database, then this option is obviously not be available.
* The predictive models are automatically re-trained and applied by a server component, which requires an additional server installation (SAP Predictive Analytics Server / SAP Predictive Analytics Factory). It is easiest to have both installed together on a Windows Server.
* Predictive models can be trained either in the web interface of the Predictive Factory or in the locally installed SAP Predictive Analytics Desktop. In this example we are using the Desktop installation.
* For a detailed list of the supported technical environments, please see the Product Availability Matrix (PAM) for SAP Predictive Analytics. The latest version of the PAM can be found on the SAP Predictive Analytics Community. Currently SAP Predictive Analytics 3.2 is supported on SAP HANA 1.0 SPS12 revision 122.04+ and SAP HANA 2.0.

Optional components:

* Strictly speaking it is optional, but to get the full value of SAP Predictive Analytics you really ought to install the Automated Predictive Library (APL) on SAP HANA. These are the algorithms that will train the predictive models directly in SAP HANA. Without the APL the data would be temporarily extracted by SAP Predictive Analytics.

Data Preparation

To save time and effort we want to leverage the existing logic of your SAP BW system. This is achieved by exposing the SAP BW objects as SAP HANA Calculation Views, using the SAP BW functionality you might already be familiar with. In the properties of an Advanced DataStore Object for example, just select the option “External SAP HANA View”.

Many different SAP BW objects can be exposed as SAP HANA View. SAP NetWeaver 7.5 SPS09 for instance allows to generate SAP HANA Views for the following objects: SAP HANA-optimized InfoCube, DataStore object (classic), DataStore object (advanced), InfoObject, query as InfoProvider, CompositeProvider, local CompositeProvider in BW Workspace, Open ODS view as PartProvider of a CompositeProviders and query.

On activation you will see the corresponding Calculation View in SAP HANA’s “Content” structure.

 

At this point we can connect with SAP Predictive Analytics to the data. If your data is already structured for the predictive purposes, that would be great. Most likely some restructuring will be needed though. If you want to predict a customer’s behaviour for example, then each of your customers must be described in a single row of data. This condensed view on the level of customer is needed so that a predictive rule can be found that assigns predictions to each customer. Chances are that currently this information is spread across multiple objects and rows.

Such restructuring can be done using the “Data Manager” component in SAP Predictive Analytics. Data is restructured on the fly without needing a permanent duplication. The interface is graphical to support efficiency. Scripting is possible also if desired.

Joining the Calculation Views is straight-forward.

 

You will appreciate “Data Manager” most when having to aggregate multiple rows to achieve the level of granularity need for the prediction. So sticking to our example, you will need to turn the multiple rows of a customer into a single row that is probably much wider.

A customer’s sales history could be turned from thousands of rows / transactions into one line with columns such as

* revenue previous quarter
* revenue two quarters ago
* change of revenue between quarters in Euro
* change of revenue between quarters in Percent
* …

 

Creating these aggregates by hand, ie with SQL syntax, would be possible, but not scalable. Very quickly you can have hundreds if not thousands of variables. You could create a similar revenue analysis for the last 4 years and just looking at quarterly aggregation alone creates 16 columns. These can be created individually for each of your product segments. Having only 10 segments, would lead to 160 columns. Think of all the other information that could be helpful. Don’t exclude anything on personal gut feel.

Also multiple aggregation types can be used at the same time. It could be helpful to know what was the maximum revenue per quarter, or the minimum, or how many transactions occured each quarter. Other aggegration types are possible of course. Go as broad as possible to keep the chance that some of these variables will be helpful. Later on, the automated predictive framework can identify and select which variables are actually needed.

These aggregates are created graphically in Data Manager. Again, this is purely semantic. There is no data transfer or duplication.

 

Train Model

Now that the data is structured accordingly, let’s create our predictive model! Using the Automated Mode of SAP Predictive Analytics, this is like taking a photo with a digital camera. Focus on the picture you want to take and press the shutter. Here you can focus on the business question in mind, provide the view of your data and let the predictive framework do it’s work to find the best possible model.

If you are not a Data Scientist, this automation opens up the world of Predictive Analytics / Machine Learning for you. If you are a Data Scientist, this automation can take care of the recurring tasks, saving you a tremendous amount of time. So your back is covered and you can explore and experiment with new use cases.

 

You will know that the model is trained directly in SAP HANA, reducing complexity and avoiding the usual caveats of data extraction (data consistency, delay, possibility of errors, security…).

When the model has been found, some overall statistics are presented to you. I will not explain the details here, but the overall quality is represented in two easy to understand measures:

* Predictive Power
* Predictive Confidence

If you really want to know more about these statistics and what happens in the background, please have a look at this blog: How does Automated Analytics do it? The magic behind creating predictive models automatically

 

The model is ready to go into production. However, you may also be interested in the inner workings of the model. Very often these can reveal interesting insights. The default representation is easy to understand and can be shared and discussed with a larger audience. A specialist user has more options available to investigate the predictive model.

 

Apply Model

We can use this model immediately to create predictions. Or for long-term productive use, have it maintained by the Predictive Factory. The Factory can create predictions when needed for your processes.

 

These new forecasts can be written directly into SAP HANA. You can choose how previous forecasts should be handled at this point. They can either be overwritten so that only the latest forecast is kept, or you can keep all historic forecasts to build up an overview of how the forecasts have changed over time.

 

Maintain

It can also retrain the predictive model as and when required. As the underlying data changes, any predictive model will deteriorate over time. Therefore models need to be consistently monitored and retrained. The Predictive Factory takes care of this, also collecting a history of how the quality of the model has been changing over time.

 

Combine with SAP BW Content

Now that the predictions are kept in SAP HANA, it is easy for SAP BW to pick up these forecasts and join them with existing content.

 

This example uses an Open ODS View:

 

We have seen how SAP BW can be enriched with predictions, leveraging existing logic and content. Any application can now access this information to improve decisions. This could be an automated process that has improved visibility for automated decisions. Or, to name a second example, it could be a Dashboard, which is now showing both the past as well as predictions in a single view.

The next blog in this series is looking at the typical steps of a project. http://bit.ly/2zwvBDt #SAP #SAPCloud #AI