Neural networks meet least squares Monte Carlo at internal model data

Abstract In August 2020 we published “Comprehensive Internal Model Data for Three Portfolios” as an outcome of our work the committee “Actuarial Science” German Actuarial Association. The data sets include realistic cash-flow models outputs used proxy modelling life and health insurers. Using these data, implement hitherto most promising model in modeling consisting ensembles feed-forward neural networks compare results with least squares Monte Carlo (LSMC) polynomial regression. To date, latter represents—to best knowledge—the accurate function productively use by insurance companies. An additional goal this publication is a more precise description other researchers, practitioners regulators interested developing solvency capital requirement (SCR) models.