Data Science and Technology Applications
Data Science and Technology Applications
This domain tests the use of data, statistical models and technology in actuarial work, with emphasis on governance and reliable inference.
Core rules and concepts
- Data quality covers completeness, accuracy, consistency, timeliness and granularity.
- Supervised models predict an outcome; GLMs offer interpretable risk factors, while machine learning can capture non-linearity.
- Split data into training, validation and test sets; avoid leakage from future or target information.
- Overfitting gives good in-sample fit but poor out-of-sample performance.
- Explainability, privacy, cyber controls and model governance are part of actuarial responsibility.
Worked example
A claim-frequency model uses policyholder age, vehicle group and region. The team fits a Poisson GLM and a gradient-boosted model. The boosted model has lower training error but similar test error and is harder to explain. They retain the GLM for pricing, use the boosted model as a challenger, and monitor both against actual claims. This separates predictive gain from governance cost.
Traps
Judging models only on training accuracy, leaking target data into features, ignoring class imbalance, and treating a black-box score as an explanation. Also assuming more data always improves quality.