An individual model for claims reserving based on Bayesian neural networks

Following the approach in W¨uthrich (2018), we propose a new approach for individual claims reserving and we show how individual development factors can be modelled as the prediction target of a system of Bayesian neural networks. This approach allows to take into account the complete information on policyholders available to the insurance company and to provide a new application of Bayesian neural networks to obtain a stochastic claims reserve. This contribution will show a case study that compares the individual chain ladder approach and the Bayesian neural networks model.