Effective Statistical Learning Methods for Actuaries I : GLMs and Extensions
(2019) ISBN: [9783030258191], 441 pages, published
Files
ISBARP_2019_40_denuit_EffectiveStatisticalLearningMethodsforActuariesI_GLMsandExtensions_oa_dial.pdf
Restricted Access - Adobe PDF
- 8.79 MB
Details
- Authors
- Abstract
- The present material is written for students enrolled in actuarial master programs and practicing actuaries, who would like to gain a better understanding of insurance data analytics. It is built in three volumes, starting from the celebrated Generalized Linear Models, or GLMs and continuing with tree-based methods and neural networks. After an introductory chapter, this first volume starts with a recap’ of the basic statistical aspects of insurance data analytics and summarizes the state of the art using GLMs and their various extensions: GAMs, mixed models and credibility, and some nonlinear versions, or GNMs. Analytical tools from Extreme Value Theory are also presented to deal with tail events that arise in liability insurance or survival analysis. This book also goes beyond mean modeling, considering volatility modeling (double GLMs) and the general modeling of location, scale and shape parameters (GAMLSS). Throughout this book, we alternate between methodological aspects and numerical illustrations or case studies to demonstrate practical applications of the proposed techniques. The numerous examples cover all areas of insurance, not only property and casualty but also life and health, being based on real data sets from the industry or collected by regulators. The R statistical software has been found convenient to perform the analyses throughout this book. It is a free language and environment for statistical computing and graphics. In addition to our own R code, we have benefited from many R packages contributed by the members of the very active community of R-users. We provide the readers with information about the resources available in R throughout the text as well as in the closing section to each chapter. The open-source statistical software R is freely available from https://www.r-project.org/. The technical requirements to understand the material are kept at a reasonable level so that this text is meant for a broad readership. We refrain from proving all results but rather favor an intuitive approach with supportive numerical illustrations, providing the reader with relevant references where all justifications can be found, as well as more advanced material. These references are gathered in a dedicated section at the end of each chapter. The three authors are professors of actuarial mathematics at the universities of Brussels and Louvain-la-Neuve, Belgium. Together, they accumulate decades of teaching experience related to the topics treated in the three books, in Belgium and throughout Europe and Canada. They are also scientific directors at Detralytics, a consulting office based in Brussels. Within Detralytics as well as on behalf of actuarial associations, the authors have had the opportunity to teach the material contained in the three volumes of “Effective Statistical Learning Methods for Actuaries” to various audiences of practitioners. The feedback received from the participants to these short courses greatly helped to improve the exposition of the topic. Throughout their contacts with the industry, the authors also implemented these techniques in a variety of consulting and R&D projects. This makes the three volumes of “Effective Statistical Learning Methods for Actuaries” the ideal support for teaching students and CPD events for professionals.
- Affiliations
Citations
Denuit, M., Hainaut, D., & Trufin, J. (2019). Effective Statistical Learning Methods for Actuaries I : GLMs and Extensions (Springer Actuarial). Springer. https://doi.org/10.1007/978-3-030-25820-7