Schedule of Machine Learning & Artificial Intelligence
| Day 1, Tuesday, 24 November | |
| 16:00 - 18:00 | What actuaries should know about tree-based machine learning models By MAHY Samuël, MARECHAL Xavier |
| Day 2, Thursday, 3 December | |
| 16:00 - 18:00 | Vizualization and Interpretation of Machine Learning Models : Making Machine Learning Models Explainable By HAINAUT Donatien |
| Day 3, Tuesday, 8 December | |
| 16:00 - 18:00 | Boosting insights in insurance tariff plans with tree-based machine learning methods By ANTONIO Katrien |
-
From 16:00 to 18:00
What actuaries should know about tree-based machine learning models
By MAHY Samuël, MARECHAL Xavier- What is artificial intelligence and statistical machine learning?
- Model performance, over fitting and cross-validation
- Classification and regression trees
- A dummy example
- Regression tree algorithm
- Pruning and parameter tuning
- Bagging, random forest and boosting
- An example in R
-
From 16:00 to 18:00
Vizualization and Interpretation of Machine Learning Models : Making Machine Learning Models Explainable
By HAINAUT Donatien- Global interpretation with variable importance measures and partial dependence plots.
- Local analysis with local interpretable model agnostic explanations (LIME).
- The SHAP model for local interpretation.
- The LIME approach is illustrated with the R library "lime".
Abstract:
Machine learning (ML) models are often considered as “black boxes” due to their complex inner-workings. More advanced ML models such as random forests, gradient boosting machines (GBM), artificial neural networks (ANN), among others are typically more accurate for predicting nonlinear, faint, or rare phenomena. Unfortunately, more accuracy often comes at the expense of interpretability. This training focuses on recent approaches to interpret machine learning models. We first review global interpretation methods including partial dependence plots (PDP), Individual Conditional Expectation (ICE), Accumulated Local Effects (ALE), permutation feature importance, leave-one-covariate out (LOCO), H statistic and global surrogate models. The second part of the module explores local approaches as the local interpretable model-agnostic explanations (LIME) and Shapley values (SHAP). The course is illustrated with R (package IML) and concepts are applied to interpret a neural network fitted to a credit default data set (package KERAS). The R code shall be provided participants.
-
From 16:00 to 18:00
Boosting insights in insurance tariff plans with tree-based machine learning methods
By ANTONIO Katrien- Apply the tree-based machine learning methods in frequency and severity modeling
- Practical implementation: choice of loss function, (hyper)parameter tuning, model evaluation
- Assess model insights and compare resulting tariff plans
- Construct interpretable surrogate models starting from black box models.