Telematics combined actuarial neural networks for cross-sectional and longitudinal claim count data (Q6556599)

The paper provides novel cross-sectional and longitudinal claim count models for vehicle insurance. The architecture of these models is based on a combined actuarial neural network (CANN), that combines classical actuarial models with a neural network. This approach takes advantage of the main features of classical models, as well as the flexibility and the capacity to capture intricate relationships provided by the neural network. After presenting two datasets, that is the contract dataset and the telematics dataset, the CANN claim count models are discussed in depth and then applied to specific dataset. Model performance is extensively analyzed.