robert brooks, pwc
TRANSCRIPT
Using big data and predictive analysisRobert Brooks & Matthew Tomlinson
www.pwc.com
PwC
Agenda
The Background
The Key Requirements
What is it?
What is needed to make it work?
The ApplicationHow have we used in the past?
PwC
Background
What is Data and Predictive Analytics?
Data mining
Future probabilities and trends
3
PwC
Data, analytics, technology and information management are all evolving at a rapid pace that is set to accelerate in the future… and it will spare no industry
1980’s
1990’s
2000’s
2010+
1970’s
Reports
OLTP
Punchcards
Data Processing
DecisionSupport
Websites
Audio
Finance Management
Analyst
Modelf(x)
X1
X2
X3
Y1
Y2
MultivariateAnalysis
BusinessIntelligence
PredictiveModeling
InformationWorker
Simulation &Visualization
SocialMedia
The DataScientist
EmbeddedAnalytics
Mobile
The DataWarehouse
The DataWarehouseAppliance
Big Data
RDBMS
Smart Phones & Tablets
Increasing pace of evolution
BackgroundAdvances in Data & Analytics over time
Access to a large wealth of modelling algorithms and
techniques
Cheap(er) storage and computing power (e.g. cloud
based solutions)
Exponential development of data available (internal and external to organisations)
A significant change in paradigm:
4
PwC
BackgroundPolicing data
5
of staff records
1,000s
of
addresses
millions
of victimsmillions
of ANPR hitsbillions
of vehicle records
100sof phone records
100,000s
of financial records100,000s
of offender records100,000s of witness statements
millions
of intelligence reports
100,000s
of calls
millions
of crime reportsmillions
PwC
BackgroundInternet of ThingsConverging and connected technology…
6
Smart devices
Sensors
Biometrics
Wireless Connectivity
Nanotechnology
Analytics
Robotics
• A multi-trillion dollar emerging industry
• 50 billion connected devices by 2020, generating 40k exabytes of data
• 54% of global top performing companies are investing more in sensor technologies
• Identified by WEF as a phenomenon that will dramatically transform economic activity (including insurance)
Wearables
Sources: PwC Digital IQ survey, IDC, Business Insider, World Economic Forum
Data storage
PwC
BackgroundCreating the internet of…everything!
7
*50 billion connected devices by 2020, generating 40k exabytes of data
Smart sensors & connected devices everywhere*
PwC
BackgroundWhat is predictive modelling?
• Using past data to find patterns
• Most well known applications is credit scoring
• Statistical models used to segment areas to together
• Principally using GLM (generalised linear modelling)
• Evolving data science towards algorithmic Machine Learning
• Who
• When
• What
• To which group should we …
8
Predictive models Questions
PwC
BackgroundTypes of machine learning
9
Supervised Learning: pre-labelled data trains a model to predict new outcomes
Example: Sorting LEGO blocks by matching them with the colour of the bags
Unsupervised Learning: Non-labelled data self organises to predict new outcomes (e.g. clustering)
Reinforcement Learning: feedback to algorithm when it does something right or wrong
Example: Child gets feedback ‘on thejob’ when it does something right or wrong
PwC
Model
Testing !Outcome
Action
BackgroundGeneral process
PwC
Key requirements
What is needed to make it work?
The question you are try to answer
Data
Tools and systems
11
PwC
People
Culture
Senior buy-in and support
Ensure clear communication
Ensure outputs are simple and easy to interpret
Skillset
Processes
Identifying the right individuals
Establish training
Collaboration including experts in other areas
The Key Requirements
Systems
PwC
Response
Integrate with existing processes
Keep the output simple
Understand the limitations
Calculation
Key variables and correlation
Business and expert judgement and challenge
Ethics on using personal data
The Key Requirements
People Processes Systems
PwC
Software
Consider users
Start with a proof of concept
Consider open-source
Data
Merging multiple datasets
Align with other analytics/ business intelligence
Consider sources: Direct, Indirect and External
The Key Requirements
People Processes SystemsPeople
PwC
The Application
How have we used in the past?
15
PwC
The ApplicationPredictive models: Professional Gamblers
What’s the problem?
Tighter regulation and smaller profit margins require betting companies to be more selective about their customers.
How we helped?
16
Identify the customer
Determine the cut-off
Understand the customer
PwC
The ApplicationPredictive models: Predictive Asset Maintenance
What’s the problem?
A power company needs to reduce the amount of network downtime from assets that fail.
How we helped?
17
Highlight assets with a high risk of failure
Integrate with existing maintenance schedule
Use real-time data feeds
PwC
The ApplicationPredictive models: Talent retention
What’s the problem?
A media company wanted to understand and manage the loss of talent in the organisation.
How we helped?
18
Predict those at high risk of leaving
New performancemanagement system
Targeted interventions
PwC
The ApplicationPolicing
Questions?
This publication has been prepared for general guidance on matters of interest only, and does not constitute professional
advice. You should not act upon the information contained in this publication without obtaining specific professional advice. No
representation or warranty (express or implied) is given as to the accuracy or completeness of the information contained in this
publication, and, to the extent permitted by law, PricewaterhouseCoopers LLP, its members, employees and agents do not
accept or assume any liability, responsibility or duty of care for any consequences of you or anyone else acting, or refraining to
act, in reliance on the information contained in this publication or for any decision based on it.
© 2016 PricewaterhouseCoopers LLP. All rights reserved. In this document, “PwC” refers to PricewaterhouseCoopers LLP
which is a member firm of PricewaterhouseCoopers International Limited, each member firm of which is a separate legal
entity.
Robert BrooksT: 020 7212 2311M: 07725 [email protected]
Rob Brooks FIAAssociate Director,Actuarial Services
Matthew TomlinsonT: 0117 309 2538M: 07843 [email protected]
Matthew TomlinsonSenior Associate,Data Assurance & Analytics