ACT 401 · Year 4 · Semester 1 · 3 credits · AI & Machine Learning

Predictive Analytics for Insurance (Exam PA)

From raw policy records to board-ready executive strategy: master the statistical learning, predictive modelling, and persuasive communication of SOA Exam PA.

Full practice exam · 30 Q · 120 min

Season 1 · 8 episodes

The Pricing Defense: Actuarial Predictive Analytics in the Hot Seat

When an adverse selection crisis drives Apex Mutual's flagship commercial auto line into a 118 percent combined ratio, newly appointed lead predictive actuary Maya Lin must build, validate, interpret, and defend a complete modern predictive rating system before the state insurance commissioner and the board of directors.

Protagonist · Maya Lin, FSA, MAAA, a rigorous, mathematically sharp actuarial data scientist recently promoted to lead the Predictive Analytics Unit at Apex Mutual.

Setting · The underwriting floor and executive boardroom of Apex Mutual in Chicago, alongside the formal hearing rooms of the Illinois Department of Insurance.

Stakes · If Maya cannot engineer a high-performing predictive model that passes strict ASOP standards and regulatory scrutiny within eight weeks, Apex Mutual will suffer a forty million dollar underwriting loss and be forced to exit commercial auto, putting hundreds of local policyholders and two hundred staff at risk.

Pick an episode

Episodes

Season finale: Maya delivers a regulator-approved hybrid predictive rating structure—pairing an elastic-net regularized GLM with gradient boosting insights and full SHAP interpretability—that slashes loss ratios by fourteen points while complying completely with ASOP 12, 41, and 56.

From the lecturer: Welcome to ACT 401. SOA Exam PA and modern predictive analytics demand a dual fluency: mathematical precision in statistical learning and crisp, executive-grade communication. As you engage with these lessons, keep RStudio and a structured workbook open beside you. Do not treat these videos as passive lectures; pause to inspect data distributions, fit models alongside the derivations, and critically evaluate model diagnostics before the solutions are revealed. Every topic in this course bridges the gap between pure computer science machine learning algorithms and regulated actuarial decision-making—covering GLMs, regularized regression, tree-based ensembles, and explainable AI techniques like SHAP values and partial dependence plots. Pay equal attention to the qualitative framing and executive report structures, as clear communication of model limitations and business impacts is what distinguishes an elite credentialed actuary. Treat every project as if you are presenting directly to the Chief Actuary and C-suite stakeholders.

Cold open

Seven thirty Monday morning. Maya arrives at Apex Mutual to find Chief Actuary David Mercer staring at a loss ratio dashboard flashing deep red. Competitors using machine learning have cherry-picked Apex's best commercial fleet risks, leaving Apex with a catastrophic combined ratio of one hundred and eighteen percent.

preparing lecture…

Transcript

Translate ambiguous commercial mandates into tractable predictive analytics tasks
Define the appropriate unit of analysis, exposure offset, and target variable while preventing data leakage
Align statistical loss functions and performance metrics with commercial key performance indicators and actuarial standards of practice