Radical Raters XI - by Neil Fleming

Rocket Science Meets Bayesian Models: Space Risks – Small Binary Data.

Space Bayes Presn - Neil Fleming.pdf

Download PDF • 1.26MB

Neil Fleming's presentation outlines a data-driven approach to assessing satellite launch risk for space insurance underwriting. Neil outlines how we can predict binary outcomes (success/failure) of launch vehicles 6–36 months into the future.

This presentation delves into how we can develop models with predictive capability that always produce valid results. These models are evaluated on training fit, predictive accuracy, and temporal sensitivity using metrics such as LOOIC, Lift, and Gini. In this way, Neil demonstrates how we can select the best-performing models to guide underwriting decisions.