You are on page 1of 12

Advanced Data Analytics

White Paper

Table of Contents
1.0 Introduction

2.0 Data Analytics

2.1 Challenges

2.2 Concepts and Capabilities

2.3 Evolution/Road map

3.0 Implementation

3.1 Methodology and Framework

3.2 Reference Architecture

3.3 Platform Offerings

4.0 Use Cases

4.1 Commercial

4.2 Government

11

5.0 For more information

12

1.0 Introduction
For most organizations today, the rapid growth of data
that must be managed represents a daunting challenge
that is pushing the limits of their existing IT infrastructure.
At the same time, driven by increased use of a wide
range of new technologies, the composition of this data
is rapidly shifting. Industry analysts estimate that about
80% of this new data is unstructured, file-based data that
cannot be analyzed using traditional database tools.
To address this big data challenge, organizations need
the highly skilled data scientists, the tooling, and the
domain expertise, to unlock the value of their data
assets. Further, organizations require a method to
access these resources without incurring the expense
of recruiting and building an entire analytics team.
Unisys provides a comprehensive solution to help
organizations address your unique data analytics
challenges. Our Advanced Data Analytics solutions
combine a comprehensive array of data analytics
tooling with enterprise scale disciplined engineering
approaches from Unisys that are tailored to your specific
needs. Our solution combines a proven analytics
platform with skilled data scientists to partner with you
to solve your most challenging analytics problems.

Business Insight

Human
Insight

Visualizations &
Business Intelligence

Transformations & Views

Expressive
Analytics

Effective
Information
Management

Raw Data

2.0 Data Analytics


Data analytics cover a wide range of techniques to derive
actionable information from available data. The traditional
approach, Business Intelligence, focuses primarily on
retrospective analysis, i.e. descriptive information of
prior activities and trends. However, modern analytics
techniques provide the capability for both predictive
analysis and can be applied to a wider variety and
higher volumes to data to accurately forecast trends,
identify significant events and underlying causes to help
better manage your business and compete effectively
through a culture of data driven decision making.

However, achieving a culture of data driven decision


making is not a straightforward task. Organizations face
many challenges in effectively using the data available
to them to achieve their goals. These include:
Too much data to process Advances in Information
Technology (IT) and consolidation of systems and
architectures provide access in vast amounts of data.
Not enough trained staff to analyze data Only keeping
up with established reporting requirements.

Data Processing & Storage


Storage

Our unique consumption-based model allows organizations


to right-size their analytics projects to meet their
needs. No longer are effort stymied by lack of skills,
technology, or IT infrastructure. Furthermore, customers
pay for what they use, not what they may use. This
allows for analytics efforts to follow the more natural
and agile course of iterative knowledge discovery.

2.1 Challenges

Analytics
Machine
Processing

Using our proven approach, Unisys experts augment


and enhance existing IT capabilities to provide analytics
over multiple platforms and devices. With our end-to-end
data analytics solution, our customers make timelier,
well-founded decisions that allow the organization to
accelerate success and overcome the big data challenge.

Efficient Data
Processing

Lack of understanding of the importance of the data


Typical business analysts focus on specific types and
scopes of data. They dont have the skills and experience
to understand the volumes and varieties of available data
and their analytical potential.

Inability to access and process data Enterprise and


infrastructure lack connectivity and capacity to provide broad,
yet secure data access.
Lack of analytics tools Investigating, selecting, acquiring and
integrating analytics tools requires specific expertise.

Business Intelligence Even in advanced analytics


implementations, traditional BI techniques can still
contribute when applied to structured data such as the
following:
-- Data Warehouse

Access to potential users via mobile devices Providing


secure access to data analytics when and where its needed.

-- Data Marts

Unable to correlate disparate data sets such as social


media, asset management, events, etc. Data comes to
the organization in many formats and levels of structure.
Traditional analysis methods and tools cant effective ingest
and correlate it.

-- SQL

Budget limitations to create new capabilities need to share


data and analytics across organization.
Unisys, leveraging our experience and expertise in advanced
data analytics has proven techniques and approaches to
help you overcome these challenges to effectively exploit
the potential of your organization through data analytics.

2.2 Concepts and Capabilities


To achieve the maximum benefit of advanced analytics,
Unisys approach encompasses a holistic view of Enterprise
Data Management, including all of the following items:
Data Integration efficient operations and high
performance necessitates managing data effectively.
Activities applied to facilitate integration include:
-- Data Migration
-- Data Consolidation
-- Data Profiling
-- Data Cleansing
-- Data Provenance
-- Master Data Management
Data Visualization Visualization is an effective tool
to enable users and analysts to identify relationships
and data features quickly and accurately. Our analytics
platform includes this features for use in visualizing
aspects of your data, as appropriate:
-- Drill-down Reports
-- Dashboards
-- Geographical Reports
-- Social Networks

-- RDBMS
Data Intelligence In addition to the data itself, there is also
valuable information that can be derived from metadata and
data interrelationships. These techniques include:
-- Link Analysis
-- Entity Resolution
-- Entity Extraction
-- Business Rules
Data Governance In order to assure the integrity and
accuracy of any analytical results and also to maintain
compliance with all applicable internal and regulatory
handling and usage controls and constraints, Unisys
maintains a comprehensive governance approach using
the following techniques:
-- Data Stewardship
-- Data Auditing
-- Data Access
-- Data Ownership
-- Data Quality
-- Data Security
Through our comprehensive approach to data management,
Unisys provides a solid foundation to build out an
effective advanced data analytics capability/platform.

2.3 Evolution/Road map


Meeting the challenges of increasing volumes and varieties
of data and exploiting the capabilities of advanced
data analytics on the route to becoming a data-driven
organization encompasses both process maturity
and data complexity as illustrated by the following:

Modeling and
Forecasting

Extend

Classification

/A

Machine Learning
Simulation

at
eg
r
In
t

Process Maturity

Global
Optimization

Big Data & NoSQL

Business
Intelligence & Data
Warehousing

Google

Hadoop

Big Table

Map/Reduce

Backward-looking
(Descriptive)

Leverage for
large-scale
analytics and data
mining

Linear
Programming

bs
or

Pattern
Recognition

Advanced
Data
Analytics

Scale-out

Forward-looking
(Predictive)

Data Scientist Group

SQL
RDBMS

ETL

STAR
Schema

OLAP

Splunk

Hive

Dynamo
MongoDB

Cassandra

EMC

Leverage for largescale application


development &
information
management

Greenplum

HBase

Current Client Environment


Low
Volume, Variety, Velocity

Analytic Tools Integration

Multi-TB Turning Point

High
Volume, Variety, Velocity

Data Maturity
Data Complexity As volume, velocity and
complexity increase, traditional structured data
repositories become less effective. Need to move
toward big data and NoSQL repositories.
Process Maturity As demands from data analytics
evolve from backward looking (descriptive) analysis
to forward-looking (predictive) and ultimately to
prescriptive analysis, application of advanced analytics
and data science is required to recognize and
extract non-obvious relationships and patterns.
Together, data and process demands require advanced
data analytics methods and tools to provide the
leverage for large-scale analytics and data mining.

As your demands process maturity and requirements


for data complexity grow, our offerings provide a range
of services and service levels appropriate to your
needs. Our Advanced Data Analytics experts assess
your available data, then develop and utilize custom
Hadoop-based infrastructures that address both
technical and logistical challenges. Common steps
include solution development according to technical,
legal, security, and other requirements; data ingestion
from multiple sources and formats; data cleaning,
organization, processing, and transformation; and
predictive modeling and algorithm development.

Process Maturity

Data Input As illustrated in in the figure, our platform


includes raw data inputs for open source and internal
sources using Flume and Pentaho for ETL operations.
In addition, Scoop provides and Enterprise Data
Warehouse (EDW) ETL capability for structured data
input. Data are loaded into Hadoop for storage,
providing a massively scalable data lake that can
incorporate structured and unstructured data including
documents, emails, blogs, presentations, and images.
BDAP also incorporates a RESTful API web service
for direct data ETL connections and automation.

Unisys Advanced
Analytics
for IoT
Wearables
Sensors
Unisys Advanced
Analytics
Predicitve
Perspective
Unisys Service
Prescriptive
Excellence office
Customer Service
Management
Service
Data Analysis
Management
Predicitive
Analytics
SNC Performance
Service
Analytics
Management
Analytics
Descriptive
BMC
Truesight
Manual
Reporting
Excel
Data Complexity

3.1 Methodology and Framework

Our platform will also accommodate partitioning


structures to provide access-controlled repositories
for secure compartmentalization of particular data
sets. We can also implement fine-grained access
controls and statistical metrics, and an automated
source catalog utilizing the capabilities provided in
Storm, Hbase, Accumulo, Hive and Cascalog.

The Unisys Analytics platform is illustrated in the


Figure below. It provides a full suite of tools for the
implementation of advanced analytics capability.
Features of our platform are described below.

Data Discovery Data Discovery allows files to be


tagged by features including owner, and type while
also providing customized tags that can be created to
align to specific objectives. Our platform incorporates

3.0 Implementation

Ingest
Multiple
Sources

Manual
Discovery &
Understanding

Use R to create
multiple views
from tables

Stitch all views together


to create de-normalized
columnar tables

Monitor, Measure,
Analyze and Improve
Refine and
Improve

Create Data Products


into Production

- Implement Real Time


Production
- Update Infrastructure
- Integrate with other
Applications

Check Models

Feedback

Apply Linear Regression


and Clustering

Refine and
Improve

Load multiple
views into Hadoop
for analysis

Define Variables

Analytic Design and


Data Exploration
Modeling and
Analytics
Implementation

Validate de-normalized
columnar table
definitions

Decision Point

Create Models
Predictive Analytics
Statistical Analysis
Using Tools such as R, SAS

BI Analytics
Dashboards
ETL Analysis
Data Warehouse
Using tools such as
Tableau, Splunk, etc.,

an algorithm layer that provides Data Science and an


analytics toolbox to analyze, extract required details
and insight into the underlying data. The following
machine learning packages are part of our platform:
Entity Extraction
Sentiment Analysis
Key Word Extraction
Concept Tagging
Taxonomy Classification
Data Preparation Routines
In addition to the machine learning package mentioned
above, Unisys also exposes its complete machine learning
toolkit in order to enhance and develop new algorithms.
The machine learning toolkit also includes: Naive Bayes,
Network Graphing, and Support Vector Machines. These
algorithms can be used to build predictive advanced
analytics and prescriptive reporting for decision support.

Our interface layer provides the ability to extract


results either via Web Deployed Applications or
programmatically through exposed APIs. It also provides
a visualization layer that integrates D3 libraries as
well as standardized analytic reports. We will also
develop specific APIs for users and applications in
the Secure Cloud providing access both through
the Hadoop file system and SQL interfaces.
Data Security The Unisys platform incorporates
enhanced auditing functions including detailed
reports. This includes enriched data tags and
features to enable feature and data access controls
and specified user roles and data policies.
Agile Engagement Model Unisys approach to
understanding your data and working with you to
develop and implement effective and appropriate
analytics capabilities are based on Agile development
principles as illustrated in this Figure.

Proof of concept - Data Rationalization

Identify
Data
Sets

Analyze
Data
SMEs

Define
Data
Products

Refine
Predictive
Models

Refine and Improve

Proof of concept 4 to 6 weeks


Limited Amount of Data - One or two ideas

Deployment to Production

Validate
Analytic
Results

Identify
Production
Data

Integrate
Analytic
Engine

Monitor
Refine
Measure

Refine and Improve

Production 6 weeks
Per Data Product

During the proof of concept phase, we work with


your staff to identify available data and get an overall
understanding of the business needs and objectives you
want to accomplish. Working interactively, we develop a
proof of concept model limited to one or two questions
or concepts that represent your needs. From this model
we work with you to refine and validate both the model
and the analytic results to your satisfaction. When were
mutually comfortable with the results and capability,
we move into the production phase. In production, we
continue the agile engagement process to expand and
formalize the analytical framework incorporating additional
data and predictive models while continuing to monitor
and refine the accuracy and the accuracy and integrity of
the results. We find that this becomes an ongoing process
as once Clients become aware of the initial potential
afforded by advanced data analytics, they recognize new
questions to ask and predictions to be made from their
data. Unisys will continue to support this interactive
process toward making you a data-driven organization.

OLAP &
What-if

Dashboards
& Reports

3.2 Reference Architecture


Unisys advanced data analytics reference architecture
is illustrated below. It is divided between the
analyst-focused services and functions and the underlying
IT-focused infrastructure for data management.
This layered structure enables Unisys data scientists and
architects the flexibility to tailor our analytics platform to your
specific needs while providing the capability for a clear and
cost-effective expansion and upgrade path in the future to
accommodate increased capacity and new technologies.

3.3 Platform Offerings


The following exhibit illustrates our Advanced Data Analytics
as a Service Environment. It provides a full-featured
analytics environment, based on proven configurations
and best of breed technologies, that can be easily
and quickly deployed and hosted on industry leading
infrastructure providers, including Amazon Web Services,
Microsoft Azure and our own Forward! environment.

Search &
Retrieval

Forecasting
& Modeling

Geospatial

Visualizations & Business Intelligence

Analyst
Focused

Ad-Hoc
Analysis

Precomputations
& Aggregations

Data Mining

Modeling &
Simulation

Analytics

Information
Extraction

Consolidated
Views

Data Enrichment

Transformations & Views

Environment for development of


data products and discovery

Analytical Processing and


model execution

Data refinement and global data


views; metadata management

IT Focused

Data Security

Aggregates

Authorization
Data
Acquisition

Cleansing &
Standardization

Indicies

Raw
Data

x-formed
Data

Encryption

Data Processing & Storage


Infrastructure

Network

Labeling

Disk

Processor

Data ingestion
and labeling;
preservation of
source data lineage

Unisys Advanced Data Analytics Platform


Ingest Data

Open Data
Social Data
Logs
Serv. Mgmt.
Asset Mgmt.
Videos
Audio
Pictures
Docs

Data Strategy
Data Integration
Data Security
Storage Rationalization

Normalize
Transformation of data
at scale
Integration of algorithms,
Taxonomies and/or
3rd Party data
Reformat, Visualize
Data Enrichment

Productize

Actionable Intelligence
Data Mining
Machine Learning
Predictive Models

powered by

Advanced
Data Analytics
Sealable Platform &
Data Scientists

Customer Data &


Domain Expertise

Our platform combines the flexibility of a configurable


architecture, the agility and resiliency of cloud-based
infrastructure, the power of an integrated set of
industry-leading tools and the stability and reliability of a
configuration proven and refined with multiple clients.
We also support our analytics platform with consulting
services Unisys Data Scientist as a Service.
Our experienced staff can support you with:
Increased understanding of hardware and software
toolsets to support data analytics
Business justification for follow-on work with Consulting
engagements defined use cases aide in building the
business justification
Understanding of Unisys differentiation in the Big Data
and Data Analytics market place through introduction of
related IP
We can also help you lay out a data analytics strategy,
including the key initiatives needed to put data
analytics into action in your organization and targeting
the to-be state for analytics in the organization. We
provide a roadmap to go from raw data to business
insight. A typical 3-6 month engagement includes:

Fraud Detection
Volume
Forecasting
Sentiment
Analysis
Ticket
Optimization
Customer 360 0
Recommender
Engine
Loan Delinquency
Reservation
Optimization
Survival
Hardware Model

Business Insights &


Mission Effectiveness

Assessment of current information capabilities


Identification of desired information capabilities
Description of target
High-level roadmap
Executive briefing of results

4.0 Use Cases


4.1 Commercial
Predictive Sentiment Analysis Measurement of user
and customer sentiment is both vital to business success
and challenging to accomplish accurately and effectively.
Traditional measures, such a Likert scale surveys provide
easy to digest objective data, but frequently miss
subtleties in attitudes and topics outside of the survey
ratings. However, analytics in the form of predictive
sentiment analysis can be used on unstructured data,
such as written comments elicited from users and
customers to automatically determine overall attitude
without the necessity of manually reading and tabulating
comments. We have applied this technique to service desk
response surveys as illustrated in the following graph:

Pharma customers was able to increase their


volume forecasting from 20% to 90% which
translated it in a 30% increase efficiency

From these data, we can determine areas such as Microsoft


Outlook which while not being a significant source of tickets,
has the highest sentiment scores, indicating an area
where it would be highly beneficial to concentrate efforts
to reduce tickets overall and expedite resolution of the
ones that still come in. Other surveys have shown dramatic
differences in sentiment expressed in written comments
versus objective scale ratings due to the respondents being
able to detail their specific concerns without criticizing
individuals or parts of the process that are working.
ITSM support for Pharma Unisys provides IT Service
management for one of the worlds largest biotechnology
companies. Unisys service desk and desk-side support,
aligned with their business operations, has supported
and sustained 34,000 desktop/laptop/tablet systems
and other services (service desk, telework support) at
approximately 45 locations in 40 countries worldwide.

10

Unisys provides centralized desktop/laptop/tablet IT


services, including application delivery, modernization,
and compliance with Amgen security policies. In our
continual effort to optimize the services provided to
Amgen, we have achieved a cost savings of $8.8M.
To support continuous improvement in our delivery of
services, we needed to analyze multiple documents,
categorize and link the different profiles to better assist end
user support roles. The goal was to automate the process
of identifying the correct topics from the text documents
and build a knowledge graph that would allow different
categories of the support team to know what the issue
was without spending the upfront diagnostic time. We
integrated the required data (unstructured and structured)
in our analytics platform to serve as our data lake. We
leverage the following algorithms from our analytic platform:
Text Extraction
Concept Tagging
Sentiment Analysis
Knowledge Base and Inference Engine

Unisys worked directly with the client to identify the different


hypothesis, requirements, business values and results. With
our interactive analytic approach we enhance and updated
the approach and results as needed. Many of the challenges
were due to the complexity of the analysis including:
Analyzing text and documents is challenging task and
a number of ingestion algorithms had to be applied to
extract data as required
Accuracy of the learning algorithm improves but in many
cases additional data are needed to explain variance and
requires time and data investment
Developing appropriate roles and persona is a
business question that many clients dont fully
understand themselves
As a result, we were able to improve the overall
operational efficiency of the support team and improve
communication and sentiment of the end users.
Telco Data Products Library Unisys provides
analytic services to a major Telco in the
form of a data products library.
For this engagement, we developed several
model of their operations, providing measureable
business value. These models include:
Customer Segmentation and 360 views of customers
In this model we use multiple internal databases to create
a 360 view of each customer. This provides our client
with the ability to categorize and group customers based
on past behavior and historical trends.
CPU Utilization and Batch Scheduling In this model
we analyze log files from transactional databases to
identify historical usage and be able to predict spikes.
This enables our client to examine usage by specific
customer on a minute-by-minute basis to define and
classify behavior and be able to predict future activity.
They can also perform more exhaustive stress testing on
the online systems to predict the likelihood and impacts
of high utilization.

Routing optimization Through this model we identify


customer behavior while leasing numbers with specific
routing to predict underutilization of routing options
which translate into decreasing revenue. It also
enables the client to identify trends and patterns on
routing information to offer new features to customers
in the future.

4.2 Government
GSA Product Data Library For the GSA, Unisys also
maintains a product data library which includes these models:
Churn Analysis and Opportunity Growth Unisys leverages
transactional sales data for the PSC codes that FAS
manages in order to identify and prioritize customers
who are more likely to decrease their GSA spend. We
also developed a Customer Flight Dashboard to assist
in prioritization of customer retention. Customer Survey
Results (i.e., customer loyalty & detractor feedback) data
are matched up to each level as practicable.
Customized Segmentation Validation We enable GSA/
CAR to group customers that share key characteristics
into clusters, to better enable FAS to target marketing and
sales efforts. Apply exploratory data mining techniques
(clustering) from data according to similar buying patterns
and then contrast those groups to existing categories and
segmentation that are being defined by the CAR team.
Opportunity Chatter Unisys provides GSA with the
ability to perform text analysis over FBO.gov to provide
an understanding of keywords coming through different
listings from FBO.gov. These keywords are ranked based
on observations of the keyword within the FBO data and
coalesced with the market category data and other PSC
and NAICS information. Data provided enables reference
back to actual opportunity listing from FBO.gov.
CRM and Integration of Data Unisys developed concepts
and uses of the CRM data in conjunction with the
transactional data to understand the relationship between
buyer purchasing overall customer satisfaction.

Telco customer is capable now to offer new


online products specifically targeted to group
of customers with an 90% effectiveness

11

USDA Loan Delinquency Mitigation To identify and


develop ways to reduce default rates for mortgage loans
across portfolios, Unisys developed a two stage model to
identify the probability that a loan would be delinquent and
estimated the outstanding portfolio balance of delinquent
loans, providing the ability to proactively manage and
mitigate loan loss and the individual and portfolio levels. As
illustrated in the figure below, the accounts and outstanding
balance that make up a transition state of interest can
be identified and then grouped by their probability of
delinquency for mitigation and remediation actions.

5.0 For more information


To learn more about Unisys advanced data
analytics capabilities and offerings, please
go to: http://www.unisys.com/offerings/
application-services/big-data-analytics.

For more information visit www.unisys.com


2015 Unisys Corporation. All rights reserved.
Unisys and other Unisys product and service names mentioned herein, as well as their respective logos, are trademarks or registered
trademarks of Unisys Corporation. All other trademarks referenced herein are the property of their respective owners.
Printed in the United States of America

12/15

15-0560

You might also like