Kernel Poisson regression machine for stochastic claims reserving

This paper shows how the kernel Poisson regression machine (KPRM) can be applied in the context of claims reserving. The paper concentrates on the chain-ladder technique, within the framework of the chain-ladder linear model. It is shown that the KPRM can provide stable reserve estimates. The ordinary cross-validation (OCV) and the generalized cross-validation (GCV) functions are introduced to determine hyperparameters which affect the performance of the KPRM. Experimental results are then presented which indicate the performance of the KPRM.