Posts Tagged ‘predictive’

SAP Discusses Latest BI Roadmap

August 15th, 2013 3 comments


Yesterday, many SAP BusinessObjects customers, mentors, partners and even a few competitors listened in as Steve Lucas as he unveiled SAP’s bold, new analytics strategy.  The twittersphere as active as tweets were tracked using #allaccessanalytics, we had a lot of fun.  Zimkhita Buwa quipped:

Steve emphasized that SAP isn’t delivering a ‘mission accomplished’ banner.  It’s just a ‘mission’ banner.  “Where we are and where we are going.”  There is a bunch still to come…  One of the first bold things he did announce was that the personal edition of SAP Lumira is free and you can download it now.  In addition you can also register for free access to the SAP Lumira Cloud.  You can register for free here.

New Mindset

It was clear to hear Steve’s passion for analytics.  His passion goes way, way back… and is always welcome.  In fact, I think this is an old picture of Steve’s car back in 2003.

He talked about BusinessObjects founder Bernard Liautaud and the rich legacy from which this SAP BusinessObjects BI Suite comes.  Steve took a note from Visha Sikka to say that SAP is and will innovate and that the Innovator’s Dilemma is crap.  Customers will continue to see great new innovations from SAP.

The new mindset from SAP delivers three pillars:

  • Enterprise Business Intelligence – For the Entire Organization
  • Agile Visualization – For the Business
  • Advanced Analytics – Data Science for Everyone

Enterprise Business Intelligence

With 60,000 customers, SAP continues to have to largest market share within the Business Intelligence space.  SAP BusinessObjects didn’t invent Business Intelligence but our experience in the space is very rich… and if not SAP BusinessObjects then who?  SAP is going to continue to build out capabilities for the Enterprise organization.  This is squarely focused on our SAP BusinessObjects product line.

At the same time the market has shifted and everyday business users want to be able to connect to their data quickly and easily and get new insights and share those insights with colleagues… and perhaps not have IT involved at all.  That brings us to agile visualization.

Agile Visualization

In SAP’s mind, Agile means it’s incredibly easy to adopt and deploy.  It should be light-weight.  Visualization means that its high quality and there is storytelling behind it.  The visualization tells a story.  It gives new insights.

So is this just Lumira?

No.  Steve was clear about this.  This is not just Lumira.  Lumira and Lumira cloud are a part of the agile visualization strategy, but the agile visualization strategy.  He promised that more would be shared at TechEd, etc.  I think this is great news!  We are beginning to see integration today between Lumira Cloud and on-premise systems and it sounds like this will continue.

Advanced Analytics

Steve said that this is more than just predictive.  As with Lumira, Predictive is a part, but there’s more.  SAP’s view is that this is not just for data scientists but data science for everyone.  It sounds we might begin seeing more ‘smart’ functionality build into the analytics.  We’ve already seen predictive leveraging the interface of Lumira for easy data access.  I can definitely envision a lot of possibilities here.

Key SAP Executives

Michael Reh (@reh_michael) – is leading development.  Passionate about analytics.

Christian Rodatus (@crodatus)  – Go to market executive for analytics.  18 years at Teradata.  Brings big data perspective.

Shekhar Iyer (@shekharlyer111) is leading the BI organization.  Brings a predictive perspective.

Jack Miller (@jackmillerIII) is Chief Customer Officer – in charge of generating successful, happy customers.

Jayne Landry (@jaynelandry) is Crystal Kiwi, working closely with Shekhar.

Other Highlights

SAP BusinessObjects BI Suite 4.1 is out of ramp-up.  All the KPIs were hit, so I would expect a GA release soon with the release of SP1.

There was a great video of a WebIntelligence-like product that was running on top of HANA.  It was written completely in HTML5.  I’ve never seen this before.

Screenshot of HTML5 WebI-like Tool

There was another nice video of Lumira doing geospatial analysis using the new Lumira Visualization Extensions which were released with SP11.  Timo Elliott recently did a nice blog post talking about this topic.

Lumira with Geospatial Capabilities

Lumira with Geospatial Plug-in

On September 9th, SAP is planning to launch a new BI online support site.  It looks as if they are following in the footsteps of the HANA launch site.  They briefly showed a mock-up of what it might look like.

Magic Bus

Steve revealed the new bus.  Yes, it’s a literal Big Data Bus.  SAP will be rolling out a mobile briefing center that will be used to showcase SAP’s latest and greatest. I think it’s one of SAP’s ways of saying there’s  50 ways to leave your… niche BI tools.   So hop on the bus Gus!

I couldn’t resist.

My Thoughts…

If you are an SAP customer and haven’t yet purchased SAP BusinessObjects, there is no better time than now.  The integration between SAP and the BusinessObjects BI Suite is second to none.  Here are a list of just a few of the unique advantages you can leverage when reporting against SAP ERP and SAP BW with SAP BusinessObjects:

  • Support for SAP CTS+
  • Integration with Solution Manager
  • Support for RRI (Report to Report Interface)
  • Support for BICS (3x faster than legacy BAPI interface)
  • Best Heterogeneous data source support
  • Best Slice/dice performance within MS Excel
  • Embedded Analytics within SAP Business Suite EP6
  • Crystal Reports options for Advanced List Viewer (ALV)
  • Semantic layer support for Infosets, ABAP functions and ABAP Queries
  • 100% In-memory support for all your SAP data

If your organization is committed to SAP Business Suite, then leveraging SAP BusinessObjects to provide reporting off those solutions is a no brainer.

Secondly, have a look at Predictive Analysis.  Although this product is relatively new, SAP has come a long way very quickly.  SAP has combined the core self-service Lumira (Visual Intelligence) product together with the power of R to deliver world class predictive analytics to the data analyst.  The interface is extremely easy to use and if you haven’t seen it, check out the post I did where I provided a product walk-through.  It may not necessarily replace SAS today, but it can deliver tremendous value by shortening the length of time it takes data analysts to build, model and run predictive algorithms.  Users are no longer wholly dependent on the small number of statisticians to provide predictive  and statistical analysis.  Predictive Analysis is a game changer.

Thirdly, get familiar with SAP’s simplified licensing.  Back in the day, when BusinessObjects was just one product, licensing was easy. Over the years as the BusinessObjects BI portfolio has grown, not everyone was ready to leverage the new technologies such as WebIntelligence, Dashboards, Explorer, etc.  As a result, BusinessObjects allowed customers to buy products à la carte to keep the pricing competitive.   A lot has changed.  Today, Business Intelligence is ubiquitous.  Everyone needs it and organizations should want to leverage the same solution for multiple types of users who have different analytic needs.  Back when the only product was had was Crystal Reports, I used to show how Crystal Reports provided enterprise reporting, adhoc reporting and dashboards. SAP’s approach was to simplify this licensing through bundles.  At the beginning of 2013, SAP offered BI Suite licensing which provided two important changes:  concurrent user licensing and a powerful software bundle of nearly every product in the SAP BusinessObjects Business Intelligence Product Suite.


This #allaccesswebinar didn’t answer all our questions but one thing was clear:  SAP is fully committed to an easy-to-adopt analytics product suite for all users that serves the enterprise through both on-premise and cloud.  They are committed to delivering solutions that: compete head-to-head against the newcomers, deliver customer value and are agile and easy to adopt and use.

If you want more information on the latest published roadmaps from SAP, go here.

Now… hop on the bus Gus!

«Good BI»

Building a Predictive Model

March 6th, 2013 7 comments

Last year SAP launched a new predictive solution for the BusinessObjects Business Intelligence Platform.  Prior to 2012, SAP had partnered with SPSS to provide predictive functionality; however once SPSS was acquired by IBM, it was time for SAP to develop their own solution.  This gave birth to Predictive Analysis.

In version 1.0, Predictive Analysis was built leveraging the Eclipse framework.  Since then, the Predictive Analysis Interface has been merged with Visual Intelligence and provides customers with a  powerful predictive analytics and visualization solution.

When Do I Use Predictive?

Predictive Analysis should be used anytime you need leverage a predictive algorithm to get additional insights based on statistical modelling.  Here is the wiki site which talks about predictive analysis in more detail.

Since Utilities is my primary area of focus, here a quick post I wrote about the different Predictive Use Cases for Utilities.  Predictive Analysis is widely used in most industry verticals for different use cases.  It is probably most widely used in retail as organizations look for ways of increase their share of the customer wallet.

(Here is more  information on my BI decision tree.)

How to Build a Predictive Model

Using Predictive Analysis is extremely easy.  Organizations can easily access corporate data and use that corporate data to gain new insights that allow them to decrease cost and increase profitability and efficiency.

In the case, I would like to see how I might predict the propensity of my transformers to failure.  I will look at historical and current asset data together with failure rates and see if I can predict which assets are most prone to failure.

Connecting to the Data

With Predictive Analysis you can access any data from any of your corporate or local data sources.  In my case I’ll be using a local Excel spreadsheet which contains the key failure data I’ll need for this exercise.  If you’d like you can download it here:

Step 1 – Connect to Data

Although I can pull information from a Universe or HANA, I am selecting an Excel spreadsheet called Transformer Analysis.

Step 2 – Select the Excel data source

Now I can preview the data and confirm I have the correct data set.  I can choose to exclude columns if I wish to.

Step 3 – Preview and Read the data into Predictive Analysis

Once the data has been loaded, you can begin to look through the data, visualize it and analyze if for additional insights.  In our data set, the key columns we will be looking at are:

  • Status:  OK/Failure – what is the current status of this asset
  • Attribute Fields:  VegMgmt, Overloads, PM Late, Miles From Ocean, etc. – these are fields that will help us determine if we can predict future failure
Step 4 – Now we have full capabilities to Manipulate and Visualize the Data

Building the Model

Next we will want to begin to do some analysis or we can move directly into Predictive Analysis.  Since this is not a demo of Visual Intelligence, I will simply move directly into predictive mode by changing my perspective and selecting Predict:
Step 5 – Change Prespective from Prepare to Predict

Now that I am in the Predict perspective, I can select from any number of prebuilt functions.  A short overview of what these functions are used for can be found here.  Here is a screenshot of those functions:

PreBuilt Predictive Functions (Click for full size)

 For this predictive exercise, we will use a Decision Tree.  We will use the attributes of the data to see what influences or predicts the likelihood of failure.  Double click on the R-CNR Tree and it will be added to the workspace.

Step 6 – Select the R-CNR Algorithm

Next, hover over the R-CNR algorithm and select the properties tab.

Step 7 – Modify the Properties

The properties dialog for the R-CNR Tree algorithm will appear.  Here we want to specify which column we are trying to predict and which fields will influence that prediction.  Therefore we will specify Status as the independent column.  All the remaining columns will be influencers.  In my example I selected:   Status, VegMgmt, PMLate, Overloads, MilesFromOcean, Manufacturer, WaterExposure, MultipleConnects, Storm, AssetType, Repairs.

I also changed the name of the “Newly Added Columns” from PredictedValues to PredictedStatus .

Step 8 – Fill in the Parameters for the R-CNR algorithm

Once the values have been selected, we will be ready to run the algorithm.

NOTE:  I did find that when using the algorithm, there is a
bug in the R algorithm and the column names cannot contain a space
in them.  If they do, you will get an error when you try and run
this algorithm.

Running the Model

What’s great about predictive analysis is that as an analyst, you can build your predictive workflow one step at a time.  After each step you have the option of running your algorithm. up so that point.  You can use either the “Run Till Here” button on the individual step in the analysis process OR you can press the green arrow to run the entire workflow.

Step 9 – Run the Algorithm

After running the algorithm successfully, we will want to view the results.

Step 10 – Success!  Now Let’s look at the Results

Once we click on the results tab, we’ll be able to see any new content that’s been created.  In our case, we will see there is a new column which has been created called, PredictedStatus.

Step 11 – Seeing the Results

Now if we want to drill into the algorithm results generated by Predictive Analysis we can do that as well.

Step 12 – View the Charts

  In the case of a R-CNR algorithm, we can see the graphical decision tree which represents the algorithm.

Step 13 – Decision Tree Graphic (Click on link to see full size)

You may not be able to see the entire decision tree on your screen based on the resolution of your screen and the number of levels generated by the algorithm.  There is a limit to how many times you can zoom out in order to see the graphic.

If you would rather see the results of the algorithm in text format you can do that as well.

See Algorithm result in Text format

Analyzing the Results

Once I’ve completed all the steps in the process, we can continue to visualize data by leveraging any new columns that have been generated by the output of the workflow.  In this case, we have generated one additional column called, PredictedStatus.  This can now be used in analysis.

Click on the Visualize (which is next to the Charts button) and we will switch to visualization mode.

Step 14 – Visualize the Results (Click to see full screen)

In this case, we pulled in Status, Predicted status, and created a Count measure based off the AssetId field.  In the visualization above, we are comparing predicted failures vs. actual failures.

In this case we can see that of the assets that are currently OK, 290 of them are predicted to fail!

That’s Insight.  That’s Powerful.  That’s Predictive Analysis.

«Good BI»

Categories: BI Platform Tags:

Visual Intelligence Goes Predictive….

December 13th, 2012 5 comments

It was only a few short months ago when Visual Intelligence was released.  We were promised a rapid development and release cycle, but I never expected this many changes is such a sort time.

Here is a quick summary:

  • 1.0.1 – Visual data analysis and data discovery for the Desktop
  • 1.0.2 – Freehand SQL, Publish to Streamwork, Proxy support.
  • 1.0.3 – Universe  4.x
  • 1.0.4 – Basic Predictive, Export to Explorer (with Explorer 14.0.4), Geocoding Lat/Long
  • 1.0.5 & 1.0.6 – Bug fixes
  • 1.0.7 – Universe 3.x, 32-bit release, cell limit increase to 30M

Now both Visual intelligence and SAP’s newest Predictive Analysis are coming together and are available today in a single release.

Where To Find the New Release

The new, combined release is branded as Predictive Analysis 1.0.7.  It uses the same code base as Visual Intelligence, so any of your saved Visual Intelligence documents will open and work just fine with Predictive Analysis 1.0.7.

You can find the release under: Software Downloads > SAP Software Download Center > Support Packages and Patches > A-Z > P > SBOP Predictive Analysis Service > SAP predictive Analysis 1.0 > Entry by Component > Predictive Analysis 64-bit > Windows on x64 64bit.  The file size is 238885 kb and was released on 11/19/2012.

Download PA on Service Marketplace

If you want to install it, there are a couple of tricks to getting it to install correctly, so you’ll want to follow my instructions below.

Installing Predictive Analysis 1.0.7

The download and initial installation is very straightforward – but before long I needed installation help, which was the motivation for this blog post.

First you must uninstall the following old products:

  1. SAP Predictive Analysis
  2. Visual Intelligence
  3. R (any version)

If you don’t you’ll receive a message at installation time that says, “incompatible products installed.’

I also recommend when you install the product to use “Run As Administrator” just to make sure there aren’t initial issues.

It will complete without asking any questions and like Visual Intelligence will allow you to use it for 30 days without a keycode.

Predictive Analysis is ready to go.

Using Predictive Analysis with R

The great news with SAP Predictive Analysis is that the software distribution now includes a menu option which will use web services to connect to the R distribution library and automatically download and install R for you.  The current release supported here is R 2.15.1.

The one challenge is that the installation of R will cause an error because the R distribution has changed since SAP Predictive Analysis was released and there is an error in the install script which will cause the installation script to fail.

The problem is explained in this SAP NOTE:

You can avoid the error and everything will work fine if you fix the R install script prior to attempting to install it.

64-bit Fix

  1. Simply download the updated inputpackages64.txt file here (Right Click – Save As): 
  2. Replace the existing inputpackages64.txt in the $ with this one.  (Default installation directory is:  C:Program Files (x86)SAP Predictive Analysis)
  3. Continue with the R Installation instructions.

 32-bit Fix

  1.  Simply download the updated inputpackages32.txt file here (Right Click – Save As): 
  2. Replace the existing inputpackages32.txt in the $ with this one.  (Default installation directory is:  C:Program FilesSAP Predictive Analysis)
  3. Continue with the R Installation instructions.
NOTE:  Although SAP software is often downloaded as a .EXE, you can always
manually unpack the .EXE file by using a utility like WinRAR.  Here is
a previous post about the topic:

Installation Help – Installing R Step by Step

From the File menu choose Install and Configure R.

Install and Configure R

This will make sure that the correct version of R is downloaded and configured with SAP Predictive Analysis.  It will download R-2.15.1

R Installation Screen

You will see the download process begin.  The R download package is almost 50 MB.

Downloading Progress...

R will begin to install after the download.

Extracting files and installing R

Next it will download individual additional libraries which are needed.

Downloading supplemental libraries

If you updated the installation file, R and all the required packages should install correctly.

The final step to complete the installation is go into the Configure tab and set the configuration to point to your R library.  The default directory should be correct.

Final Step - Tell Predictive Analytics Where R is Located

Fixing Your R Installation Manually

If you were not able to modify the installation file, you can also complete the installation of R manually.  Here is the message you are likely to see if you did not update the installation program.

ERROR:  R installation not successful.  Detail: R-2-15.1 is installed successfully. But failed to install the required R packages.  Ensure that you are connected to the internet and you have Administrator rights to install software on the computer.

Common Error When Installing R

The Good News is that the core R software has installed correctly.  The only part that failed was the downloading of all the add-on libraries that Predictive Analysis needs for all the build-in predictive functions.  We will need to install the following libraries one by one.

  • arules
  • XML
  • caret
  • DBI
  • monmlp
  • PMML
  • rJava
  • reshape
  • Plyr
  • Foreach
  • Iterators

You can manually download them one-by-one from the R website:

Or you can download all the required files from my blog and unzip them:

  1. Open the RGui Console (Start > Programs > R > Rx64 2.15.1)

    Installing Packages Manually

  2. Choose Packages > Install Package(s) from local zip files
  3. Select the zip files one by one and install them, you may ignore any warnings.  After the package installs successfully you will get the following message:  Package packagename successfully unpacked and MD5 sums checked.
  4. If you want to test that a library is installed, go to the R Console command prompt and type:  library(rJava).  Press Enter.  It is successfully installed if you don’t see any errors.
  5. You can also check in your R library folder, which is located here:  [R_INSTALLATION_FOLDER]library.  This is typically C:UsersPublicR-2.15.1library for Windows 7 users.

Additional Troubleshooting Help

If you need additional help, check out the following excellent resources:

PA 1.0.7 Troubleshooting Guide

Install Guide for PA 1.0

Let me know if you’ve got any additional installation or configuration issues and solutions in the comments below.

Hopefully I can save your a few hours of troubleshooting.


Visual Intelligence and Predictive Analysis will continue to be licensed separately.  If you only need data visualization, then you will only need to purchase a license for Visual Intelligence.  If you want to add Predictive, then you need to purchase a license to enable that functionality.

License Key Screen in Predictive Analysis

Here in the Help > Enter License Key, you can see that both Visual Intelligence and Predictive Analysis license keys are recognized.

«Good BI»