Jackson
Actuarial-Data Science Collaboration: Key Methods and Applications (303774)
*Louise Francis, FCAS, MAAA, CSPA, Francis Analytics and Actuarial Data Mining, Inc.
Keywords: data science, modeling, R code, actuarial science, insurance, GLMs, unsupervised learning, predictive modeling,
The International Actuarial Association (IAA) is an international organization of insurance actuaries. The ASTIN section of the IAA (for Actuarial Studies in Non-Life Insurance) formed a working party on big data in order to introduce the methods and techniques of data science to actuaries. The Phase 2 ASTIN Big Data Working Party was organized to introduce key Predictive Modeling methods used by actuaries. The working party emphasizes hands-on applications and examples using a publically available database and downloadable code. The R open source program was used in examples. The following are methods that the working party is introducing: • Unsupervised Learning (Principal Components and Clustering) • MARS (Multivariate Adaptive Regression Splines) • Trees • Ensemble models • In addition, deep learning neural networks, which were introduced by a separate ASTIN Working Party, is applied to the data for comparison to the other methods. • GLMs were used as a benchmark The session will provide an introduction and summary of the working party’s analysis, detail on some of the methods, and a description of the resources that will be made available publicly by the working party.