Download Introduction to Predictive Modeling

Predictive Modeling
ASHK Seminar
November 21, 2013
Welcome to the ASHK Seminar!
►
Introduction to Predictive Modeling in Actuarial Science
►
Fundamentals of Cross-Sectional Regression Modeling
►
►
►
►
►
►
Extended Cross-Sectional Regression Modeling
►
Agenda
►
►
►
►
►
►
►
Introduction to Bayesian Computational Methods
Bayesian Regression Models
Longitudinal Modeling
►
►
►
Page 2
Mixed Models
Generalized Additive Models, including Non-Parametric Regression
Fat-Tail Regression Models
Spatial Statistics
Supervised versus Unsupervised Learning
Bootstrapping, including Simulation
Bayesian Modeling
►
►
Multiple Linear Regression
Regression with Categorical Dependent Variables
Regression with Count Dependent Variables
Generalized Linear Models
Frequency/Severity Models
Time Series, including Lee-Carter forecasting
Longitudinal and Panel Data Models
Credibility and Regression Modeling
“When the winds of change are blowing some people
are building shelters, and others are building wind mills”
Chinese proverb
AGENDA
Page 3
►
Define predictive modeling
►
Discuss applications
Let’s build some wind mills!!!!
Page 4
“Change is changing”
(Peter R. Porrino, EVP and CFO, XL Group)
”I think there is a world market for about five computers”
Tom Watson, Chairman of the Board, IBM 1943
"Computers in the future may weigh no more than 1.5 tons"
Popular Mechanics, forecasting the relentless march of science, 1949
“There is no reason for any individual to have a computer in his home”
President and Founder, Digital Equipment Corp., 1977
“What do 13 guys in Seattle know that we don’t?”
Ross Perrot (EDS) at the time when he was offered to buy Microsoft in 1980
“640K [of computer memory] ought to be enough for anybody”
Bill Gates, Founder, Microsoft, 1981
“The concept is interesting and well-formed, but in order to earn better
than a ‘C’, the idea must be feasible”
Yale University Professor in Economics on
Fred Smith’s (founder of FEDEX) term paper on next-day parcel delivery
Page 5
Just a few definitions of Predictive Modeling…
deling is a
Predictive mo
volves the
strategy that in
lection of a
creation or se
tempt to
model in an at
ible
ss
po
project the
ated with a
ci
so
outcomes as
given action.
(Wikipedia)
)
(wisegeek.com
deling is a …
Predictive mo
t
nique to predic
statistical tech
of
rm
fo
…a
future behavior
nology that
ch
te
g
in
data-min
yzing historical
works by anal
ta and
and current da
odel to help
generating a m
outcomes.
predict future
(Gartner)
Page 6
deling is the
Predictive mo
ch a model is
process by whi
en to try to
created or chos
best predict
of an
the probability
outcome
deling is a …
Predictive mo
athematical
collection of m
find] a
techniques [to
relationship
mathematical
ependent”
between … “d
dependent”
variable and “in
to
insert them in
variables to ..
p to
ical relationshi
the mathemat
values of the
predict future
target variable.
(SAS)
Really, what is Predictive Modeling?
“Predictive
“Predictive modeling
modeling is
is aa process
process to
to create
create aa statistical
statistical model
model of
of future
future behavior”
behavior”
Society
Society of
of Actuaries
Actuaries (SOA)
(SOA)
Predictive Modeling View
Traditional View
►
Set of tools, processes and
applications that gather,
integrate, query, analyze and
report on information
►
Focuses on known data that is
well understood
►
Utilizes traditional business
measurements / metrics for risk
assessment
►
Page 7
Does not include emerging
technologies/capabilities for
advanced risk assessment
►
Utilization of advanced data
mining and advanced statistical
techniques
►
Focuses on mathematically
measuring unknown
relationships between data sets/
elements that cannot be easily
managed
►
Provides ability to visualize and
explain results, and integrates
them into the business decisionmaking and operational workflow
Is there an opportunity for the actuaries?
Page 8
Why Predictive Modeling is NOW?
Industry
Industry leaders
leaders are
are aggressively
aggressively pursuing
pursuing alternate
alternate means
means to
to gain
gain aa competitive
competitive advantage
advantage such
such as
as
Predictive
Predictive Modeling
Modeling and
and advanced
advanced analytics
analytics
Use of Predictive Modeling- U.S Life Insurance Market
(2011)
Using PM
12%
Not yet
42%
Considering
46%
Key Drivers
►
Earnings demands
►
Management emphasis on profitable growth
►
Customer relationship focus
►
Pricing pressure
►
Technology innovations
►
Proven approach
Source:
Market research (Gen Re 2011)
Predictive
Predictive Modeling
Modeling
Page 9
Why Predictive Modeling IS NOT NOW?
While
While GI
GI achieved
achieved considerable
considerable success
success in
in using
using PM
PM for
for pricing,
pricing, the
the nature
nature of
of life
life insurance
insurance products
products
(long-term
(long-term duration,
duration, low
low frequencies)
frequencies) makes
makes Life
Life Insurance
Insurance aa less
less attractive
attractive target
target
Key Inhibitors
►
Lack of management familiarity
►
Size of initial investment
►
Limited business understanding
►
Enormous data preparation effort
►
Lack of skills
►
Modeling capability and capacity
►
Proof of accuracy
►
Implementation challenges
Hesitacy of implementation- U.S Life Insurance Market
(2011)
Other
25%
Size of initial
investment
20%
Lack of
management
familiarity
55%
Source:
Market research (Gen Re 2011)
Page 10
The evolution of Enterprise Intelligence
HIGH
Optimization
Enterprise Intelligence
What is the best outcome?
Prediction
►
Optimization, Decision Analysis
►
Predictive Modeling, Forecasting
Complexity
What will be happening?
What is happening now?
Analysis
Reporting
►
Statistical Analysis and Data Mining
Regression analysis
►
Time series analysis
►
Dashboards, Scorecards, Metrics, KPIs
►
Multidimensional analysis, Visualization
►
What happened
Page 11
Predict results that lie in the future
Monitoring
Why did it happen?
LOW
►
Business Value
Presentation title
►
Exploratory data analysis
►
Discover trends and patterns within data
Query, ad-hoc reporting
HIGH
Application #1: Predict Employee Performance
Oakland A’s
Page 12
►
In 1999 Oakland’s A baseball team was last
in standing, one of the last in player salary,
which lead to the departure of three stars
they could not afford to keep
►
Team’s GM - Billy Beane, assisted by a
recent Harvard graduate Paul DePodesta,
developed models to predict baseball player
performance that contradicted scouts, who
formed an opinion based on instincts and
superiority of their experience
►
Statistical analysis had demonstrated
that on-base percentage is a better
indicator of offensive success vs. speed, and
is cheaper to obtain
►
In 2002 and 2003 A’s were in the play-offs
with $41M salary (vs. $125M NY Yankees)
and remained a top-5 team thereafter, while
being last in salary for another several years
Application #2: Predict Response to Marketing Campaigns
Cigna
Page 13
►
Cigna is a $29B global health services firm
with $53B in assets. Distributes directly and
though affinity partners and brokers.
Employs 35,800 and is a #103 on the
Fortune 500
►
Customer Value Management group
(CVM) analyzes “big data” from affinity
partners to select the right telemarketing
targets and identify the right marketing
strategy
►
Analytics improves response rates,
determines the precise profile based on
personal data, and suggests a contact
strategy with tailor-made offers to optimize
results
►
Provides insights that telemarketers rely on
to extend the right offers to the right
customers at the right time
Application #3: Analyze Agency Performance
Westfield Insurance
Page 14
►
Westfield - $1.4B in GWP, 2,200 people, 1,200
agents
►
Senior leaders needed to gain a better
understanding of the agents performance
►
Transformed their analytics capability with
Analytics Resource Center (ARC) and
extended self-service analytics capabilities to
everyone in the organization reaching ~40% of
the staff use advanced analytics tools
►
Established capability to evaluate and manage
performance across its agency network
►
Made data much more easily accessible,
encouraging users to base their decisions on
hard evidence rather than intuition
Application #4: Predict Fraud
Infinity Insurance
Page 15
►
Infinity is a $1.2B motor insurer with 2,000
employees and 12,500 independent agents
►
Was interested in 1) speeding the settlement of
claims that did not contain elements of fraud
which required ability to identify fraudulent claims;
2) making better use of its investigative staff; and
3) reducing its high monthly costs for outsourced
subrogation
►
Infinity implemented predictive analytics
solutions, immediately reducing claims payments
and improving customer service
►
The accuracy of identifying fraudulent claims has
doubled,
►
Reduced SIU referral time from 45–60 days down
to 1–3 days
►
Achieved 403% return on investment from
reduction in claims payments and enhanced
subrogation
Application #5: Integrated Solution
Nationwide Financial
Page 16
►
Nationwide is a $23B insurance and financial
services company with $160B in assets. 36,000
employees and 3,500 agents. A+ by A.M. Best #108
on the Fortune 500
►
Focused on growth, they invested in building a DW
with 100Tb of user data (10x printed U.S. Library of
Congress
►
Customer Knowledge Store integrates customer,
product and externally acquired data for a single
view of the customer
►
Financial Performance Management provides a
single, integrated data management and reporting
environment
►
Goal State Rate Management allows product,
pricing and underwriting functions to access and
analyze the same data to make informed decisions.
►
Revenue Connection delivers easy-to-read
dashboards and on-demand reports to Nationwide
agents and field management
A word on tools
Page 17
Presentation title
EY | Assurance | Tax | Transactions | Advisory
About EY
EY is a global leader in assurance, tax, transaction and
advisory services. The insights and quality services we
deliver help build trust and confidence in the capital
markets and in economies the world over. We develop
outstanding leaders who team to deliver on our promises
to all of our stakeholders. In so doing, we play a critical
role in building a better working world for our people, for
our clients and for our communities.
EY refers to the global organization and/or one or more of
the member firms of Ernst & Young Global Limited, each
of which is a separate legal entity. Ernst & Young Global
Limited, a UK company limited by guarantee, does not
provide services to clients. For more information about
our organization, please visit ey.com.
© 2013 Ernst & Young
All Rights Reserved.
This publication contains information in summary form and is therefore
intended for general guidance only. It is not intended to be a substitute for
detailed research or the exercise of professional judgment. Neither the
Ernst & Young China practice nor any other member of the global Ernst &
Young organization can accept any responsibility for loss occasioned to
any person acting or refraining from action as a result of any material in
this publication. On any specific matter, reference should be made to the
appropriate advisor.
www.ey.com