2019/09/04 by Anne-Sophie Krah, Krah, Anne-Sophie, Zoran Nikolić +3
Computer Science · Mathematics · Social Sciences · #Bayesian Methods and Mixture Models #FOS: Computer and information sciences #FOS: Economics and business #Insurance, Mortality, Demography, Risk Management #Machine Learning (stat.ML) #Methodology (stat.ME) #Risk Management (q-fin.RM) #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.1909.02182
openalex publication_date 2019/09/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Under the Solvency II regime, life insurance companies are asked to derive\ntheir solvency capital requirements from the full loss distributions over the\ncoming year. Since the industry is currently far from being endowed with\nsufficient computational capacities to fully simulate these distributions, the\ninsurers have to rely on suitable approximation techniques such as the\nleast-squares Monte Carlo (LSMC) method. The key idea of LSMC is to run only a\nfew wisely selected simulations and to process their output further to obtain a\nrisk-dependent proxy function of the loss. In this paper, we present and\nanalyze various adaptive machine learning approaches that can take over the\nproxy modeling task. The studied approaches range from ordinary and generalized\nleast-squares regression variants over GLM and GAM methods to MARS and kernel\nregression routines. We justify the combinability of their regression\ningredients in a theoretical discourse. Further, we illustrate the approaches\nin slightly disguised real-world experiments and perform comprehensive\nout-of-sample tests.\n