PA Lab · SOA Predictive Analytics
Concepts you can scroll, poke, and play with.
Interactive learning materials for SOA Exam PA — built from the two-day course I teach at the Institute of Actuaries of Korea. Each module takes one idea you'd normally meet as a formula, and lets you watch it happen instead.
Decision trees, visually
Watch housing data from two cities grow into a tree: one line becomes a question, questions become branches, and a perfect training score meets its first unseen data. Splits, recursion, and overfitting — no equations.
Bagging & random forests soon
Why does averaging many overfit trees beat one careful tree? Bootstrap samples, feature randomness, and variance — as an animation, not an acronym.
GLMs: the actuary's home ground soon
Link functions, offsets, and weights — connecting what you already do in pricing to what Exam PA asks, with living residual plots.
EDA & the bias-variance trade-off soon
Train/validation/test splits and the trade-off behind every modeling decision — the module that should arguably come first, built last because it's the hardest to make visual.