Outline

Part 1

  • Overview of the lectures
  • Fundamental statistical principles underlying the modern learning approaches
  • Insurance data specificities
  • Two working levels, technical vs commercial price list
  • Recap’ of the current GLM practice, with application to claim reserving, graduation of rates, risk classification
  • Use R session

Part 2

  • Limitations of GLM tools and the need for other techniques
  • Regularization/shrinkage for GLMs: Lasso, Ridge and related penalties
  • Extensions:
    • GAMs, penalized and local likelihood
    • double GLMs, dispersion around the pure premiums
    • GAMLSS, beyond dispersion

       with application to claim reserving, graduation of rates, risk classification

  • Use R session

Part 3

  • Introduction to tree-based methods
  • Classification and regression trees
  • Bagging
  • Random forests
  • Forward stagewise regression
  • Loss functions from exponential dispersion family
  • Gradients as working responses
  • Tree-based boosting
  • Use R session