Transcript Predictive Modeling Survey - Francis Analytics Actuarial

Francis Analytics
Actuarial Data Mining Services
Predictive Modeling Workshop
Training for development and deployment
Francis Analytics and Actuarial Data Mining
The following presentation outlines a introductory training course for actuaries, and
quantitative analysts to learn basic approaches and tools to create predictive model
solutions in Property & Casualty Insurance
www.data-mines.com
Francis Analytics
Introduction
Actuarial Data Mining Services

Predictive Modeling is an important new
development for modifying traditional actuarial
approaches to risk in P&C insurance

Actuaries and Quantitative Analysts can help
advance their company’s competitive position with a
greater understanding of analytical methods and
tools

Learning opportunities in advanced actuarial
modeling area not widely available.
Francis Analytics www.data-mines.com
2
Francis Analytics
Predictive Modeling Survey Course
Actuarial Data Mining Services

A two day hands-on workshop

Delivered by experienced practitioners

Includes case study learning approach

Includes instructions in using software and the latest
statistical methods being deployed by industry
leaders.

Custom course can be designed to fit specific needs
Participants will be able to conduct a Predictive Modeling project in their organizations
and present the business case to non-technical management
Francis Analytics www.data-mines.com
3
Francis Analytics
Course Agenda
Actuarial Data Mining Services
Session
Description
Introduction to P&C Actuarial Data
Introduction and review of actuarial data for
property & casualty insurance including basic
operational view
Introduction to Predictive Modeling Methods
Survey of Predictive Modeling approaches and
current state of the art in P&C insurance
Walk-through model development phases
Walk-through of the project process for delivering
successful predictive modeling capabilities
Focus: Data Preparation Techniques
Focus teaching on data discovery and preparation
steps used in the model development process
Focus: Mathematical Methods
Focused teaching of the key statistical algorithms
used in Predictive Modeling
Focus: Statistical Software
Focused teaching and hands on tutoring of an
statistical modeling package software
Hands-on Case Study workshop
Live end-to-end walk through of a predictive
modeling project
Validation Methods
Teaching and illustration of model validation
approaches to present to management
Industry Leading Innovation
Discussion of industry leading innovation in
development and deployment of Predictive Models
Francis Analytics www.data-mines.com
4
Francis Analytics
Case Study Approach

Actual P&C example

Sample Actuarial Dataset provided

Explore challenges throughout the
exploration/model/deployment life-cycle
Actuarial Data Mining Services
Practitioners can design a custom case study with a specific modeling
need or data-set provided by the company sponsor or organization.
Francis Analytics www.data-mines.com
5
Francis Analytics
Leading Statistical Methods Covered

Generalized Linear Modeling including Logistic
Regression

Neural Network

Classification and Regression Trees

Benchmarking with Naive Bayes

Tree ensemble method

Random Forest

Support Vector Machines
Francis Analytics www.data-mines.com
Actuarial Data Mining Services
6
Francis Analytics
Software

Excel

Access

Free Software

R

Web based software

S-Plus (similar to commercial version of R)

SPSS

CART/MARS

Data Mining suites
Actuarial Data Mining Services
Participants will download, install, tour and model in R, a freeware
modeling tool
Francis Analytics www.data-mines.com
7
Francis Analytics
Learning Outcomes
Actuarial Data Mining Services

Working knowledge of the major statistical methods
used for Predictive Modeling

Practical knowledge of advanced data mining
techniques ready to be applied

General knowledge of Property & Casualty insurance
data and related operations

General knowledge of Predictive Modeling project
life-cycle management

Working knowledge of modeling software packages
and review of leading software solutions and
proprietary methods

The ability to manage the validation of models and
present the validation results to laymen
Francis Analytics www.data-mines.com
8
Francis Analytics
Logistics
Actuarial Data Mining Services
Item
Session Schedule
2 days
Overnight Homework as assigned
Course materials and Data Set
Berry, M. and Linoff, G. Data Mining Techniques
2nd Edition, Wiley, supplied to all students
California Department of Insurance public
(CAARP) data on private passenger auto will be
supplied to all participants as a download
Software
‘R’ statistical programming language, including
download and install instructions
Class Size
Limit of 25 participants
Customized Content
Arranged by agreement
Francis Analytics www.data-mines.com
9
Francis Analytics
Pricing
Actuarial Data Mining Services
Item
Course Fee
TBD
Course Materials
Included
Follow-up Consultation
Included
On-site/Group/Company Sponsored
Priced by custom proposal *
* Pricing for customized workshop content and delivery by agreement
Francis Analytics www.data-mines.com
10
Workshop Presenters
Francis Analytics
Actuarial Data Mining Services
Louise Francis, FCAS, MAAA
Louise Francis is the Consulting Principal and founder of Francis Analytics and Actuarial Data Mining, Inc. where
she leads data mining and related actuarial projects and engagements. Ms. Francis has introduced insurance
professionals to data mining methods both as a speaker at conferences and as an author of papers and articles
on data mining topics. Three of her papers were awarded the Data Quality/Data Technology prize by the CAS
(Casualty Actuarial Society) and IDMA (Insurance Data Management Association).
As an insurance professional, Ms. Francis has expertise in a variety of techniques for pricing and reserving
insurance products, both on a traditional and leading edge basis. She also has experience planning and
evaluating programs for claims loss cost reduction. Ms. Francis pioneered the application of data mining
technologies such as neural networks and decision trees to insurance data to identify patterns useful for
mitigating claims costs and making underwriting decisions. Ms. Francis introduced procedures for benchmarking
individual customer's data to industry targets to assess the performance of third party claims administrators and
to assist clients with cost reduction efforts.
Ms. Francis has over 20 years of experience in the actuarial profession. Prior to starting her own firm, she was
Director and Associate Actuary for CIGNA Property and Casualty where she provided actuarial expertise to the
Claims Division and ESIS, CIGNA's third party claims administrator. She was responsible for evaluating the
overall performance of the Claims Division and measuring the effectiveness of specific Claims Division initiatives
to reduce claims costs. Ms. Francis has also worked as a consultant with Towers Perrin and Sedgwick James.
Ms Francis is chair of the Casualty Actuarial Society's (CAS) Committee on the Theory of Risk, a committee
charged with the sponsorship and dissemination of research efforts on risk measurement and analysis. She
participates on an American Academy of Actuaries committee which develops standards of actuarial practice for
the actuarial profession.
Ms. Francis received her Bachelor of Arts degree from William Smith College and a Master of Science degree in
Health Sciences from the State University of New York at Stony Brook. She is a Fellow of the Casualty Actuarial
Society and a member of the American Academy of Actuaries.
Francis Analytics www.data-mines.com
11
Francis Analytics
Workshop Presenters
Actuarial Data Mining Services
RICHARD A. DERRIG, PH.D.
Dr. Derrig is President of OPAL Consulting LLC, established in February, 2004 to provide research and regulatory
support to the property-casualty insurance industry.
Prior to forming OPAL, Dr. Derrig held various positions at the Automobile Insurers Bureau of Massachusetts and at the
Insurance Fraud Bureau of Massachusetts over a twenty seven year period, retiring in January, 2004 as Senior Vice
President at AIB and Vice President of Research at IFB. During the spring semesters of 1994 and 2002, he was a
Visiting Lecturer and Research Fellow in the Department of Risk Management and Insurance at the Wharton School,
University of Pennsylvania. Dr. Derrig has been appointed a visiting scholar at Wharton for 2004 and 2005.
Prior to joining the Bureaus, he taught graduate and undergraduate mathematics at Villanova University, Wheaton
College (MA) and Brown for a total of thirteen years. He earned a Bachelor of Science degree in mathematics from St.
Peter's College as well as Master's and Doctoral degrees from Brown University.
Dr. Derrig is the recipient of numerous awards and recognitions for his contribution to mathematics and actuarial
sciences. He the author or co-author of numerous articles, review and contributions in the field.
He has lectured extensively on insurance topics to professional actuarial groups; the Casualty Actuarial Society, ASTIN,
trade associations, and law enforcement personnel; and to seminars at the Universities of Barcelona, Hamburg,
Montreal, Tel Aviv, Pennsylvania, Illinois, Texas, Minnesota, Wisconsin and others in the U.S. He was a director of the
American Risk and Insurance Association (1992-1995) and a recipient of the President’s Award (1997). He is an
academic correspondent of the CAS and a member of the Mathematical Association of America, American Statistical
Association, and the Association of Certified Fraud Examiners. He serves on the Insurance Fraud and Auto Injury Study
Committees of the Insurance Research Council and the CPCU Advisory Committee of the American Institutes. Since
1991, he has compiled an annotated bibliography of worldwide insurance fraud research that is made available at
www.derrig.com/ifrr/ifrr.asp.
Francis Analytics www.data-mines.com
12
Francis Analytics
Contact
Actuarial Data Mining Services
Ms. Louise Francis, FCAS, MAAA
Francis Analytics
215-923-1567
[email protected]
www.data-mines.com
Francis Analytics www.data-mines.com
13