ACTSC 456: STATISTICAL LEARNING IN ACTUARIAL SCIENCE (NEW COURSE STARTING FALL 2026)

In the Fall 2026 term I will be teaching a new course on statistical learning for actuarial science.

• The main reference is the (freely available) book Introduction to Statistical Learning with the R code that comes with it.


• We will however focus on problems and data sets relevant for actuarial science.


• Course Description: This course explores a broad range of statistical learning models used to analyze data in actuarial and financial contexts. It covers techniques from both supervised and unsupervised learning, with a focus on practical applications. Specific topics covered include: modeling principles and practice, advanced and regularized regression models, cross-validation, traditional classification models (logistic regression, LDA, QDA, KNN), decision trees (bagging, boosting, random forests), dimensionality reduction (PCA), hierarchical and K-means clustering, and an introduction to neural networks with applications to supervised and unsupervised learning problems. Some extensions to the above models may also be considered.


• Why should you take this course? Nowadays, actuaries are routinely expected to make use of modern statistical learning techniques in their work, especially in Property and Casualty Insurance. Accordingly, actuarial professional organizations have updated their exams and curriculum to recognize this important aspect of modern actuarial work. The material covered in this course will greatly help preparing exams SRM and PA from the SOA and MAS-II from the CAS. See also this website from the Department of Statistics and Actuarial Science, about pathways to accreditation. Also, as of the 2026/27 calendar, this course is required for the Predictive Analytics specialization.
  
  
• Tentative weekly schedule


• Pre-req: STAT 331, 371 or 373; antireq: STAT 441


• Questions? Email me at clemieux "at" uwaterloo dot ca (Christiane Lemieux, M3 4016)