View
1.329
Download
2
Embed Size (px)
Citation preview
#askSAP Analytics Innovations Community Webcast
Reimagine Predictive Analytics forthe Digital EnterpriseAugust 31, 2016
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 2
Legal disclaimer
The information in this presentation is confidential and proprietary to SAP and may not be disclosed without the
permission of SAP. This presentation is not subject to your license agreement or any other service or subscription
agreement with SAP. SAP has no obligation to pursue any course of business outlined in this document or any
related presentation, or to develop or release any functionality mentioned therein. This document, or any related
presentation and SAP's strategy and possible future developments, products and or platforms directions and
functionality are all subject to change and may be changed by SAP at any time for any reason without notice. The
information in this document is not a commitment, promise or legal obligation to deliver any material, code or
functionality. This document is provided without a warranty of any kind, either express or implied, including but not
limited to, the implied warranties of merchantability, fitness for a particular purpose, or non-infringement. This
document is for informational purposes and may not be incorporated into a contract. SAP assumes no
responsibility for errors or omissions in this document, except if such damages were caused by SAP´s willful
misconduct or gross negligence.
All forward-looking statements are subject to various risks and uncertainties that could cause actual results to
differ materially from expectations. Readers are cautioned not to place undue reliance on these forward-looking
statements, which speak only as of their dates, and they should not be relied upon in making purchasing
decisions.
SAP Analytics Innovations: Community Call Series
• Quarterly series for the Analytics community hosted by SAP Analytics
• An opportunity for you to direct the discussion, get your questions answered,
and end the session with some useful advice
• Live and interactive 90 minutes
• Connect on topics before, during, and after the call via twitter using #askSAP
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 4
Ashish MorzariaGlobal GTM Director,
Advanced Analytics
@AshishMorzaria
Greg MyersSAP Mentor
@gpmyers
Today’s Speakers
Richard MooneyLead Product Manager
for Advanced Analytics
@richardjmooney
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 6
Everything we touch… Every good we purchase…
In the New Digital Economy, Everything is Digitized and Tracked
Every transaction we conduct…
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 7
Customers Operational Margin Growth
How do you personalize each
interaction across all channels?
How do you improve your performance across
thousands of processes and decisions?
How do you create new products,
services, and business models?
The Digital Economy To Your Advantage…
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 8
Early Adopters Are Winning
In the next 10 years, 40% of the S&P 500 will no longer
exist if they do not keep up with these technology trends*
+9%Revenue
creation
+26%Market
valuation
+12%Impact on
profitability
* “The Digital Advantage: how digital leaders outperform their peers in every industry”: CapGemini and MIT Sloan
Those Embracing Digital Transformation are Outperforming
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 9
The Power of Predictive
Unlocks Big Value:the need for Predictive
68%of organizations using predictive analytics
realized competitive advantages.
60%of fraudulent transactions have stopped
using predictive.
28%reduction in customer churn rate with predictive.
• Use historical data to predict behaviors or outcomes
• Answer “what-if” questions
• Ensure employees have what they need to make
optimized decisions
• Fully leverage customer relationships with better insight
• Make meaningful sense of Big Data
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 10
SAP
BusinessObjects
Predictive Analytics
Data PreparationCreate meaningful and
reusable data sets
Automated AnalyticsReduce time and skills required
to create accurate models with
repeatable workflow
With Big DataUse Hadoop data with automated
techniques directly in Spark
Ultimate Flexibility
for AlgorithmsUse off-the-shelf algorithms or
bring specialized ones – such
as R functions
Accurate Results in Days, Not WeeksFor everyone: perfect for Analysts AND Data Scientists
Native in-memory SolutionSAP HANA optimized for on-the-fly
predictive data processing
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 12
SAP HANAReal-time in-memory predictive analytics platform
R Scripts
Execution of R scripts via
high-performing parallelized
vector based connection;
R scripts embedded as part
of overall query plan
Application Function
Library (AFL)
Application Function Library
(AFL) framework allows SAP,
partner, and customers to
develop, deploy, load, and
leverage their own advanced
analytic custom functions in
SAP HANA
Custom Open Source
R-Server
SAP HANA
Other Native
Libraries
© SAP AG or an SAP affiliate company. All rights reserved. 13
SAP HANAReal-time in-memory predictive analytics platform
R Scripts
Execution of R scripts via
high-performing parallelized
vector based connection;
R scripts embedded as part
of overall query plan
Application Function
Library (AFL)
Application Function Library
(AFL) framework allows SAP,
partner, and customers to
develop, deploy, load, and
leverage their own advanced
analytic custom functions in
SAP HANA
Custom Open Source
Accelerated predictive
analysis and scoring with
native in-database
algorithms
Predictive Analysis
Library (PAL)
SAP
Predictive
Analysis
Library
Automated
Predictive Library
(APL)
The predictive analysis
capabilities of SAP’s
Predictive automated
analytics engine
(formerly KXEN) in
SAP HANA
Automated
Predictive
LibraryR-Server
SAP HANA
Other Native
Libraries
APL: Automated Algorithms
Native implementation of automated predictive
algorithms: Regression
Clustering
Forecasting
Recommendation
Social Network Analysis
No data extraction required
Fully accessible from “Automated” and “Expert”
interfaces
PAL: Data Scientist Algorithms
Aims to supply most commonly used data
science algorithms (80/20 rule) natively
90+ natively coded algorithms (C++)
Freely mixable with APL algorithms
No data extraction required
R: Open Source Data Scientist Algorithms
8500+ algorithms available
Full support for custom coding
Requires data extraction (externalized process
to HANA)
Fully integrated development when using SAP
PA Suite license
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 14
Traditional Analytics Versus In-Memory Predictive Analytics
Predictive
Analysis
Library
Automated
Predictive
Library
R-Server
SAP HANA
Other Native
Libraries
• Create and apply models on very large datasets
within SAP HANA or in a Hadoop storage
transparently connected to SAP HANA
• Real-time predictions recommendations: integrate
predictive models into processes
• Native integration with SAP HANA for ERP and BW,
to provide in-applications predictive modeling
1. Copy data from transactional and external sources
2. Extract data from storage, convert & clean for analytics
3. Download analytical results & load into predictive
analytics application
4. Transfer predictive scoring results into database
SAP BusinessObjects
Predictive Analytics
vs.
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 15
Support for SAP HANA Smart Data Streaming
• Automated Analytics now supports HANA Spark
Data Streaming
• Generates CCL Code which can be deployed to
HANA SDS
• Smart Data Streaming Use Cases
o IOT Data for Predictive Maintenance and Quality
o Clickstream analysis for Marketing
o Connected Retail
HANA Smart Data
Streaming
Predictive Analytics
Automated
Modeller
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 17
Existing Process: entire dataset is transferredConnectivity = SQL only
FULL (Big Data)
dataset is transferred
for processingDataset BIG Datasets Dataset
Big Data
SQL Engines
(Spark SQL,Hive)010001100100
100101001011
100010010101
010011110101
010001100100
100101001011
100010010101
010011110101
010001100100
100101001011
100010010101
010011110101
Traditional Predictive Analytics
Data
Warehouse
RDBMS
Data platform…• power not being
leveraged properly
• just transfers data
Modeler..• Pulls in data, processes,
• Pulls in more data, processes…
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 18
Traditional
Application
Leverage Hadoop + Spark = big data store + application platform
Processing on a single server
Data Transfer
CPU/Memory scales dynamically
Processing on 100’s-1000’s of nodes
Hive QLSQL
Database
Native Application
Limited CPU/Memory
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 19
With Native Spark Modelling, processing closer to data in Hadoop
FULL training dataset
is transferred
No dataset transfer required!
Data platform…
• runs the Spark application
• processing close to dataNative Spark Connectivity
SAP BusinessObjects Predictive Analytics
Native Spark Modelling
Native Spark Modelling
• controls the process
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 20
Native Spark Modelling
Execute automated predictive models directly on Hadoop
using the Apache Spark engine
• Push the data intensive modeling workload to Native Spark -
Classification and Regression models supported
• Model Lifecycle management on Hadoop with RETRAIN and APPLY
• User structure and custom cutting strategy supported on Native Spark
• Real Time Scoring via Spark Streaming API
Benefits
• No data transfer – heavy lifting operations brought close to data
• Faster response times – 7 to 10 times performance gains
• Higher scalability – scale your training process with wider and data more
models
• Better utilization of CPUs – in distributed Hadoop environment
• Abstraction – Analysts can work with Big Data seamlessly
HDFS
(Hadoop Distributed File System)
Hive
(SQL)Spark SQL
Model Lifecycle Manager (Factory)
Scorer
Predictive Analytics Data Manager
In-DB
scoring
(Spark /Hive
QL)
Analytics
Dataset
Definition
Layer
Advanced
Analytics
Execution
LayerSpark
Streaming
(Java
Export)
Modeler -
Training
Native
Spark
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 21
Traditional Big Data vs. Big Data with SAP BusinessObjects
Predictive Analytics
Next?
Who?
How?
Big Data Analytics SAP BusinessObjects Predictive Analytics
Code Wizard Based Approach with GUI for End-Users
Big Data Developers
Ideal Tool for use by both a Data Scientist and a
Business Analyst OR Citizen Data Scientist
Data Scientists
Manually Deployed &
Monitored
Automated Deployment & Monitoring using
Predictive Factory
SAP BusinessObjects Predictive Analytics
Bringing The Gift of Predictive Insightto Business Intelligence
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 23
Descriptive (Business Intelligence) vs Predictive Analytics
Business Intelligence Predictive Analytics
• Who are my most valuable customers? • Who will be my most valuable customers next month?
• Who could become my most valuable customer and why?
• What are my most important products? • What will be my most important products?
• What products could become my most valuable products?
• What are my most successful promotions? • What promotions should I run?
• What promotions could be a good idea to run in the future?
• When did customer X visit my store last? • What is the chance of customer X visiting in the next 2
weeks?
• What were the most profitable products for
customers in my loyalty program?
• What products should I focus on to increase my profit from
customers in my loyalty program?
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 24
Smarter BI that goes beyond visual
analysis into insights that cannot hide
Predictive dashboards that
prescribe and can trigger actions
Reports that include reasons and
recommendations on next steps
Move from Descriptive to Predictive BI
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 25
Model deployed using
In-Database-Apply
Customer Database
Hancock, John M 38 D Y 4.2 N Y
Doe, Jane F 45 M Y 9.4 N N
Red, Simply F 18 S N 2.1 N Y
SQL Dataset w/ Scoring
Business Users can get on-the-fly
scoring without even knowing they
are using predictive algorithms
BI Artifact
(or even just a dataset)
SAP BI (3.x/4.x)
Embedded into any application
Cloud Applications (SaaS/PaaS/IaaS)
SQL
(Or any other application)
Embedding Predictive Analytics into BI Workflows
26© 2016 SAP AG or an SAP affiliate company. All rights reserved.
Hancock, John M 38 D Y 4.2 N ?
Doe, Jane F 45 M Y 9.4 N ?
Red, Simply F 18 S N 2.1 N ?
Model
NEW Data
(Current Customers)
Hancock, John M 38 D Y 4.2 N Y
Doe, Jane F 45 M Y 9.4 N N
Red, Simply F 18 S N 2.1 N Y
Hancock, John M 38 D Y 4.2 N Y
Red, Simply F 18 S N 2.1 N Y
Targeted List
(CR)
Significantly increase ROI through dataset reduction:
• Lower campaign costs by targeting those most likely to leave
• Increase response rate by targeting even more specifically on other attributes
• Increase C-Sat by not hassling loyal customers
Name Gender Age Marital Recent Activity C-Sat Renewed
Before
Predicted
Churn
Customer not expected to
churn, so don’t bother them!
Analysis
(WEBI / Lumira)
Batch scoring
© 2016 SAP SE or an SAP affiliate company. All rights reserved. 29Public
Sales andMarketing Operations
Fraudand Risk
Financeand HR
OtherSectors
• Churn Reduction
• Customer Acquisition
• Lead Scoring
• Product Recommendation
• Campaign Optimization
• Customer Segmentation
• Next Best Offer/Action
• Predictive Maintenance
• Load Forecasting
• Inventory/Demand
Optimization
• Product Recommendation
• Price Optimization
• Manufacturing Process Opt.
• Quality Management
• Yield Management
• Fraud and Abuse Detection
• Claim Analysis
• Collection and Delinquency
• Credit Scoring
• Operational Risk Modeling
• Crime Threat
• Revenue and Loss Analysis
• Cash Flow and Forecasting
• Budgeting Simulation
• Profitability and
Margin Analysis
• Financial Risk Modeling
• Employee Retention
Modeling
• Succession Planning
• Life Sciences
• Health Care
• Media
• High Education
• Public Sector /
Social Sciences
• Construction and Mining
• Travel and Hospitality
• Big Data and IoT
Solve Real Business ProblemsBy Optimizing Resources and Improving Margins
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 30
Predictive Process P
rob
lem
Ide
ntified
Bu
sin
ess
Re
su
ltsIdentify
Relevant
Variables
Aggregate
Prepare Data
Derived
Features &
Encode
Variables
Develop
ModelsDebrief models
Write Code for
Database
Execution
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 31
Value of SAP’s Predictive Automation
What SAP BusinessObjects Predictive Analytics does for automation:
Data Manager:
• Generate SQL for
• HANA
• Hadoop:
o HIVE, SparkSQL
• All major databases
Auto-algorithms:
Make this section obsolete
Auto-algorithms:
Numbers, strings, dates
Categorical, continuous,
textual
Date parts
Composite variables
(example: position from
latitude and longitude)
Auto-algorithms:
Classification,
regression, clustering,
times series, key
influencers
Link analysis,
recommendations
HANA (APL)
Hadoop (Scala)
Auto-algorithms:
All descriptive statistics
available
Key influencers,
decision trees,
segments, optimal
binning and banding
Communities
In-Database Apply:
Automated SQL
generation
Optimized with data
manager
Hadoop:
HIVE, SparkSQL,
Streaming (Java)
Pro
ble
m
Ide
ntified
Bu
sin
ess
Re
su
ltsIdentify
Relevant
Variables
Aggregate
Prepare Data
Derived
Features &
Encode
Variables
Develop
ModelsDebrief models
Write Code for
Database
Execution
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 32
The Predictive Factory
• Manage the lifecycle of predictive models created in
SAP BusinessObjects Predictive Analytics
• Automatically retrain, apply, test for deviation and
forecast your models
• Robustly embed predictive analytics at scale in
business processes
Key benefits
• Manage thousands of models easily and robustly
• Automate model refresh and application
• No scripting needed
• Multi-User, collaborative experience
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 33
Predictive Factory Features
Segmented Modelling• Take a dataset with thousands of
segments. e.g. Retail outlets, market
segments, geographies, products,
machines ….
• Build a model for one segment using
Automated Modeler. Import the Model
into Predictive Factory
• Segment the model in Predictive Factory
to build models for every other segment
with the same model parameters and
configuration
• Scalable to thousands of segments
• Supports Time Series in 3.0
External Commands• Run Data Preparation using external tools
• Run external, non PA Predictive Models
Sales
EMEANorth America
Product 1
Q1 Forecast
Q2 Forecast
Product 2
Q1 Forecast
Q2 Forecast
Product 3
APAC MEE
Build thousands of models
in a single operation
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 34
SAP BusinessObjects Predictive Analytics 3.0Simplify Next Generation UI
Streamlined Predictive User
Experience and Workflow• Modern design principles based on
Fiori UX and HTML 5 for a
completely reimagined user
experience
• Personalized, responsive and
simple user experience across
devices and deployment options
• In-app notifications
• X-Ray support for In-App
Contextual Help to ease first time
user experience
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 65
The Difference
Before SAP
BusinessObjects
Predictive Analytics
After SAP
BusinessObjects
Predictive Analytics
Answer any/all questions with
any/all data sources –
No limits!
In-database automated dataset
production -
No data movement!
Automated modeling and tuning
process -
Focus on accurate results, not
algorithms or code!
Native in-database and
application/process deployment -
Embed and consume for immediate
results!
On-going model management and
recalibration -
No rework necessary!
Days
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 67
Value for Business Users
• Take advantage of predictive analytics
and machine learning without Data
Science expertise
• Discover new insights in your data,
improving your business process
powered by predictive
Automated, Guided and Trusted Experience
Guided Analysis designed for Business Users,
featuring the power of Exploratory Analytics
New Discoveries
We guide you on your journey to find the answer
to your questions
Guided Machine
Discovery as Part of
SAP BusinessObjects Cloud
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 68
• Perform an embedded predictive
forecast in their planning model
• Predictive forecast runs a time series
algorithm on historic data in order to
predict future values considering trend,
cycles and/ or fluctuation.
• It can be leveraged to aid the planning
process using a data-driven approach.
Predictive Forecast as Part
of SAP BusinessObjects
Cloud
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 69
Detect fraud earlier to reduce financial loss
o Leverage the power and speed of
SAP HANA
o Integration into business processes
o Alert notification and management
Improve the accuracy of detection at less cost
o Minimize false positives with real-time
simulations
o Ability to handle ultra-high volumes
of data by leveraging SAP HANA
Predict & Prevent and deter fraud situations
o Detection based on rules and predictive
analytics to adapt to changing fraud patterns
SAP Fraud Management
with Predictive Analytics
Stella Predictive Analytics
• SAP BusinessObjects Predictive Analytics for Automated Analytics and rapid prototyping of our models
• Forward engineered into SAP HANA for real-time predictions using native, logistical regression model
• This approach allowed for identification of key predictors that more heavily influence a behavioral health outcome
• Run as a pilot to rapidly prototypethe concepts 8
Weeks for Pilot
99%Prediction Accuracy
“This tool will allow me to completely redesign the clinical
process and provide the right amount of care at the right
time. ” – Executive Director of Mental Health Provider
Stella User Experience
• Seamless UX integration• Allows for up to the
minute prediction on incoming jail records
• Flags important predictive factors for clinician
• Enables real time decision support for accurate resource allocation
Stella
This pilot allowed SAP Partner, EV Technologies, to assist Harris Logic through a successful SAP HANA pilot and later, into a cloud based architecture.
Phase 1 – Pilot – Stella 3.0 – Q1 2016• Develop use cases organized by cost, time to
deliver, and return on investment• Executed a migration of the needed JAVA
application components to SAP HANA• Successfully modelled the first two predictive
models and integrated into the pilot application – high utilizers and propensity to recidivize
Phase 2 – Stella 3.0 – June 2016• Full implementation running SAP HANA and
SAP BusinessObjects on AWS • Transitioned all pilot code to next-generation
Stella 3.0
Phase 3 – Stella 3.x+ - Q3 2016• Selected as strategic partner for the new 18
month roadmap• Developed use cases for remaining SAP
HANA capabilities including Text Analysis and the Spatial Engine
• Prioritized remaining use cases into release schedule
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 77
Online Resources
Key links
Roadmaps on SAP Service Marketplace http://service.sap.com/saproadmaps
SAP Community Network http://scn.sap.com/
Predictive Analytics Community http://scn.sap.com/community/predictive-analytics
30 days Trial Download https://www.sap.com/trypredictive
SAP BusinessObjects Predictive Analyticshttp://sap.com/predictive
Where to go to provide product feedback and ideas
SAP Idea Place https://ideas.sap.com
Predictive Idea Place https://ideas.sap.com/PredictiveAnalytics
Influence programs http://service.sap.com/influence
Sign up to our newsletter http://scn.sap.com/docs/DOC-66912
© SAP AG or an SAP affiliate company. All rights reserved.
Thank Youwww.sap.com/predictive
www.sap.com/scn-predictive
#sappredictive @sapanalytics
© 2016 SAP AG or an SAP affiliate company. All rights reserved. 79
© 2016 SAP AG or an SAP affiliate company. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or for any purpose without the express permission of SAP AG or an SAP affiliate company.
SAP and other SAP products and services mentioned herein as well as their respective logos are trademarks or registered trademarks of SAP AG (or an SAP affiliate
company) in Germany and other countries. Please see http://global12.sap.com/corporate-en/legal/copyright/index.epx for additional trademark information and notices.
Some software products marketed by SAP AG and its distributors contain proprietary software components of other software vendors.
National product specifications may vary.
These materials are provided by SAP AG or an SAP affiliate company for informational purposes only, without representation or warranty of any kind, and SAP AG or its
affiliated companies shall not be liable for errors or omissions with respect to the materials. The only warranties for SAP AG or SAP affiliate company products and services
are those that are set forth in the express warranty statements accompanying such products and services, if any. Nothing herein should be construed as constituting an
additional warranty.
In particular, SAP AG or its affiliated companies have no obligation to pursue any course of business outlined in this document or any related presentation, or to develop or
release any functionality mentioned therein. This document, or any related presentation, and SAP AG’s or its affiliated companies’ strategy and possible future
developments, products, and/or platform directions and functionality are all subject to change and may be changed by SAP AG or its affiliated companies at any time for
any reason without notice. The information in this document is not a commitment, promise, or legal obligation to deliver any material, code, or functionality. All forward-
looking statements are subject to various risks and uncertainties that could cause actual results to differ materially from expectations. Readers are cautioned not to place
undue reliance on these forward-looking statements, which speak only as of their dates, and they should not be relied upon in making purchasing decisions.