Innovation and Technology
Emerging Data
with Actuarial Analytics
Techniques
Applications
July 2019
Emerging Data Analytics Techniques with
Actuarial Applications
, 2
MARIE-CLAIRE KOISSI PhD, Professor Actuarial Innovation & Technology
Actuarial Science Program Steering Committee
University of Wisconsin-Eau Claire SPONSOR
HERSCHEL DAY FSA, MAAA, Associate Professor
Actuarial Science Program
University of Wisconsin-Eau Claire
VICKI WHITLEDGE PhD, Professor
Actuarial Science Program
University of Wisconsin-Eau Claire
Caveat and Disclaimer
The opinions expressed and conclusions reached by the authors are their own and do not represent any official position or opinion of the Society of
Actuaries or its members. The Society of Actuaries makes no representation or warranty to the accuracy of the information
Copyright © 2019 by the Society of Actuaries. All rights reserved.
CONTENTS
Abstract .................................................................................................................................................................... 4
Executive Summary .................................................................................................................................................. 4
Section 1: Acknowledgments .................................................................................................................................... 5
Section 2: Introduction ............................................................................................................................................. 6
2.1 DATA ANALYTICS FRAMEWORK .......................................................................................................................... 6
2.2 DATA SOURCES .................................................................................................................................................... 8
2.3 DATA EXPLORATION AND VISUALIZATION ......................................................................................................... 9
Section 3: Data Analytics Techniques ...................................................................................................................... 10
3.1 SUPERVISED LEARNING ..................................................................................................................................... 10
3.1.1 REGRESSION AND GENERALIZED LINEAR MODELS (GLMS) ............................................................ 10
3.1.2 TREES ................................................................................................................................................. 11
3.1.3 NEURAL NETWORKS ......................................................................................................................... 13
3.1.4 PREDICTIVE MODELING .................................................................................................................... 14
3.2 UNSUPERVISED TECHNIQUES ........................................................................................................................... 14
3.2.1 PRINCIPAL COMPONENT ANALYSIS ................................................................................................. 14
3.2.2 CLUSTER ANALYSIS ............................................................................................................................ 14
3.2.3 GENETIC ALGORITHMS ..................................................................................................................... 15
3.2.4 NEURAL NETWORKS ......................................................................................................................... 16
3.3 OTHER DATA ANALYTICS TECHNIQUES ............................................................................................................ 16
Copyright © 2019 Society of Actuaries
, 3
3.3.1 MARKOV CHAIN MONTE CARLO (MCMC) SIMULATION .................................................................
16 3.3.2 BAYESIAN ANALYSIS ..........................................................................................................................
17
Section 4: Emerging Data Analytic Technologies ..................................................................................................... 18
4.1 LIFE: MACHINE LEARNING TECHNOLOGIES FOR MORTALITY RATE FORECASTING ....................................... 18
4.2 HEALTH CARE: MACHINE LEARNING TECHNOLOGIES FOR HEALTH CARE CLAIMS MODELING .................... 18
4.3 LIFE / NON-LIFE: MACHINE LEARNING TECHNOLOGIES FOR RESERVES......................................................... 19
4.4 NON-LIFE: MACHINE LEARNING TECHNOLOGIES FOR CLAIM MODELING ..................................................... 20
4.5 LIFE / NON-LIFE: MACHINE LEARNING TECHNOLOGIES FOR INSURANCE FRAUD AND OTHER
AREAS............................................................................................................................................................................ 20
4.6 SOME ACTUARIAL PACKAGES IN R AND PYTHON ............................................................................................ 21
4.6.1 SOME ACTUARIAL PACKAGES IN R ................................................................................................... 21
4.6.2 SOME PACKAGES IN PYTHON WITH ACTUARIAL APPLICATIONS .................................................... 22
Section 5: Case Studies ........................................................................................................................................... 24
5.1 CASE STUDY 1: CHAINLADDER IN R .................................................................................................................. 24
5.2 CASE STUDY 2: CLAIMS FREQUENCY IN MOTOR INSURANCE ......................................................................... 32
5.3 CASE STUDY 3: MORTALITY (LIFE INSURANCE) ................................................................................................ 37
Section 6: Conclusion .............................................................................................................................................. 43
References .............................................................................................................................................................. 44
Appendices ............................................................................................................................................................. 53
A. Appendix A: R-Code for Case Study 1
............................................................................................................... 53 B. Appendix B: R-Code for Case Study
2 ............................................................................................................... 54
C. Appendix C: R-Code for Case Study 3 ............................................................................................................... 57
About The Society of Actuaries ............................................................................................................................... 60
Emerging Data Analytics Techniques with
Actuarial Applications
Abstract
Data analytics strongly rely on data and available computing tools. Recent years have seen an increase in
data availability and volume. Advanced computational methods and machine-learning tools have been
developed to handle this continuous flow of valuable information. The aim of this research is to survey
emerging data analytics techniques and discuss their evolution and growing use in the actuarial profession.
Data analytics’ applications in life and non-life insurance will also be provided.
Executive Summary
Data analytics involves a set of tools and techniques used to extract meaningful information from a dataset
(SOA, 2012). It encompasses several disciplines such as actuarial science, statistics, computer science,
mathematics, and marketing. Recent years have seen an increase in data availability and volume, leading to
an explosion in the concept of “Big Data” (AAA, 2018).
Copyright © 2019 Society of Actuaries
, 4
Actuaries rely heavily on data to perform analysis, make general inferences, inform decisions, and guide
predictions. They have a long history in conducting data analysis in areas such as underwriting, claim
management, pricing, risk analysis, and auditing (Shapiro and Jain, 2003; SOA, 2012). In the past, data
analysis was mainly descriptive and actuaries predominantly used programs such as Excel (SOA, 2012,
Appendix G) and C++ (Pauza and Bellomo, 2014). Although descriptive analytics is in use today, it now
represents an initial step in a more complex and data-driven analysis. Recent studies predict substantial
changes in the analytical tools used by actuaries and other professionals (Sondergeld and Purushotham,
2019; Guo, 2003; Wedel and Kannan, 2016).
Advanced data analytics packages (such as SAS, SPSS, Matlab, R, and Python) allow the user to extract more
information from a dataset, make a diagnostic analysis, and use non-standard models to make relevant
predictions. This paper aims at surveying emerging data analytics techniques with potential actuarial
applications.
The remaining part of the paper is organized as follows: Section 1 acknowledges the contributions of this
report’s Project Oversight Group (POG). Section 2 deals with the change in data source and volume. This
section also reviews some of the data visualization techniques available to actuaries. In Section 3, we give a
brief overview of several data analytic techniques. In Section 4, we review some applications of emerging
data analytic technologies in Actuarial Sciences. We also briefly describe some open-source data analytic
software that have grown in use among actuaries. Section 5 deals with three cases studies in which we use
open-source technologies for actuarial computational work. A commentary of the findings is presented in
Section 6.
Section 1: Acknowledgments
The authors gratefully acknowledge the significant contributions made by the members of the Project
Oversight Group. Special thanks are due to Dale Hall, SOA Managing Director of Research, and Mervyn
Kopinsky, SOA Experience Studies Actuary, for their leadership in guiding the project. The authors would
like to thank Korrel Crawford, SOA Senior Research Administrator, for her effective coordination of the
project and her help in getting this report ready for publication. The authors also gratefully acknowledge
the Actuarial Innovation & Technology Steering Committee of the Society of Actuaries for providing funding
for this project.
Project Oversight Group Members:
Han (Henry) Chen, FSA, MAAA, FCIA
Andrew Harris, ASA
Clinton Rheal Innes, FSA, ACIA
Karen T. Jiang, FSA, CERA, MAAA
Michael Cletus Niemerg, FSA, MAAA
Zhen Yuan, FSA
Copyright © 2019 Society of Actuaries
Emerging Data
with Actuarial Analytics
Techniques
Applications
July 2019
Emerging Data Analytics Techniques with
Actuarial Applications
, 2
MARIE-CLAIRE KOISSI PhD, Professor Actuarial Innovation & Technology
Actuarial Science Program Steering Committee
University of Wisconsin-Eau Claire SPONSOR
HERSCHEL DAY FSA, MAAA, Associate Professor
Actuarial Science Program
University of Wisconsin-Eau Claire
VICKI WHITLEDGE PhD, Professor
Actuarial Science Program
University of Wisconsin-Eau Claire
Caveat and Disclaimer
The opinions expressed and conclusions reached by the authors are their own and do not represent any official position or opinion of the Society of
Actuaries or its members. The Society of Actuaries makes no representation or warranty to the accuracy of the information
Copyright © 2019 by the Society of Actuaries. All rights reserved.
CONTENTS
Abstract .................................................................................................................................................................... 4
Executive Summary .................................................................................................................................................. 4
Section 1: Acknowledgments .................................................................................................................................... 5
Section 2: Introduction ............................................................................................................................................. 6
2.1 DATA ANALYTICS FRAMEWORK .......................................................................................................................... 6
2.2 DATA SOURCES .................................................................................................................................................... 8
2.3 DATA EXPLORATION AND VISUALIZATION ......................................................................................................... 9
Section 3: Data Analytics Techniques ...................................................................................................................... 10
3.1 SUPERVISED LEARNING ..................................................................................................................................... 10
3.1.1 REGRESSION AND GENERALIZED LINEAR MODELS (GLMS) ............................................................ 10
3.1.2 TREES ................................................................................................................................................. 11
3.1.3 NEURAL NETWORKS ......................................................................................................................... 13
3.1.4 PREDICTIVE MODELING .................................................................................................................... 14
3.2 UNSUPERVISED TECHNIQUES ........................................................................................................................... 14
3.2.1 PRINCIPAL COMPONENT ANALYSIS ................................................................................................. 14
3.2.2 CLUSTER ANALYSIS ............................................................................................................................ 14
3.2.3 GENETIC ALGORITHMS ..................................................................................................................... 15
3.2.4 NEURAL NETWORKS ......................................................................................................................... 16
3.3 OTHER DATA ANALYTICS TECHNIQUES ............................................................................................................ 16
Copyright © 2019 Society of Actuaries
, 3
3.3.1 MARKOV CHAIN MONTE CARLO (MCMC) SIMULATION .................................................................
16 3.3.2 BAYESIAN ANALYSIS ..........................................................................................................................
17
Section 4: Emerging Data Analytic Technologies ..................................................................................................... 18
4.1 LIFE: MACHINE LEARNING TECHNOLOGIES FOR MORTALITY RATE FORECASTING ....................................... 18
4.2 HEALTH CARE: MACHINE LEARNING TECHNOLOGIES FOR HEALTH CARE CLAIMS MODELING .................... 18
4.3 LIFE / NON-LIFE: MACHINE LEARNING TECHNOLOGIES FOR RESERVES......................................................... 19
4.4 NON-LIFE: MACHINE LEARNING TECHNOLOGIES FOR CLAIM MODELING ..................................................... 20
4.5 LIFE / NON-LIFE: MACHINE LEARNING TECHNOLOGIES FOR INSURANCE FRAUD AND OTHER
AREAS............................................................................................................................................................................ 20
4.6 SOME ACTUARIAL PACKAGES IN R AND PYTHON ............................................................................................ 21
4.6.1 SOME ACTUARIAL PACKAGES IN R ................................................................................................... 21
4.6.2 SOME PACKAGES IN PYTHON WITH ACTUARIAL APPLICATIONS .................................................... 22
Section 5: Case Studies ........................................................................................................................................... 24
5.1 CASE STUDY 1: CHAINLADDER IN R .................................................................................................................. 24
5.2 CASE STUDY 2: CLAIMS FREQUENCY IN MOTOR INSURANCE ......................................................................... 32
5.3 CASE STUDY 3: MORTALITY (LIFE INSURANCE) ................................................................................................ 37
Section 6: Conclusion .............................................................................................................................................. 43
References .............................................................................................................................................................. 44
Appendices ............................................................................................................................................................. 53
A. Appendix A: R-Code for Case Study 1
............................................................................................................... 53 B. Appendix B: R-Code for Case Study
2 ............................................................................................................... 54
C. Appendix C: R-Code for Case Study 3 ............................................................................................................... 57
About The Society of Actuaries ............................................................................................................................... 60
Emerging Data Analytics Techniques with
Actuarial Applications
Abstract
Data analytics strongly rely on data and available computing tools. Recent years have seen an increase in
data availability and volume. Advanced computational methods and machine-learning tools have been
developed to handle this continuous flow of valuable information. The aim of this research is to survey
emerging data analytics techniques and discuss their evolution and growing use in the actuarial profession.
Data analytics’ applications in life and non-life insurance will also be provided.
Executive Summary
Data analytics involves a set of tools and techniques used to extract meaningful information from a dataset
(SOA, 2012). It encompasses several disciplines such as actuarial science, statistics, computer science,
mathematics, and marketing. Recent years have seen an increase in data availability and volume, leading to
an explosion in the concept of “Big Data” (AAA, 2018).
Copyright © 2019 Society of Actuaries
, 4
Actuaries rely heavily on data to perform analysis, make general inferences, inform decisions, and guide
predictions. They have a long history in conducting data analysis in areas such as underwriting, claim
management, pricing, risk analysis, and auditing (Shapiro and Jain, 2003; SOA, 2012). In the past, data
analysis was mainly descriptive and actuaries predominantly used programs such as Excel (SOA, 2012,
Appendix G) and C++ (Pauza and Bellomo, 2014). Although descriptive analytics is in use today, it now
represents an initial step in a more complex and data-driven analysis. Recent studies predict substantial
changes in the analytical tools used by actuaries and other professionals (Sondergeld and Purushotham,
2019; Guo, 2003; Wedel and Kannan, 2016).
Advanced data analytics packages (such as SAS, SPSS, Matlab, R, and Python) allow the user to extract more
information from a dataset, make a diagnostic analysis, and use non-standard models to make relevant
predictions. This paper aims at surveying emerging data analytics techniques with potential actuarial
applications.
The remaining part of the paper is organized as follows: Section 1 acknowledges the contributions of this
report’s Project Oversight Group (POG). Section 2 deals with the change in data source and volume. This
section also reviews some of the data visualization techniques available to actuaries. In Section 3, we give a
brief overview of several data analytic techniques. In Section 4, we review some applications of emerging
data analytic technologies in Actuarial Sciences. We also briefly describe some open-source data analytic
software that have grown in use among actuaries. Section 5 deals with three cases studies in which we use
open-source technologies for actuarial computational work. A commentary of the findings is presented in
Section 6.
Section 1: Acknowledgments
The authors gratefully acknowledge the significant contributions made by the members of the Project
Oversight Group. Special thanks are due to Dale Hall, SOA Managing Director of Research, and Mervyn
Kopinsky, SOA Experience Studies Actuary, for their leadership in guiding the project. The authors would
like to thank Korrel Crawford, SOA Senior Research Administrator, for her effective coordination of the
project and her help in getting this report ready for publication. The authors also gratefully acknowledge
the Actuarial Innovation & Technology Steering Committee of the Society of Actuaries for providing funding
for this project.
Project Oversight Group Members:
Han (Henry) Chen, FSA, MAAA, FCIA
Andrew Harris, ASA
Clinton Rheal Innes, FSA, ACIA
Karen T. Jiang, FSA, CERA, MAAA
Michael Cletus Niemerg, FSA, MAAA
Zhen Yuan, FSA
Copyright © 2019 Society of Actuaries
