General Insurance Spring Conference 2023 - Neural network models
Neural network models have found successful applications in wide-ranging fields, from computer vision to natural language and generative text. Can it find a place in claims reserving? In this presentation, we will develop a reserving model step-by-step. Starting with a simple chain ladder, we gradually introduce incremental improvements, culminating with a probabilistic, mixture density, neural network on individual claims. This is a technical session, including both introductory and novel concepts, conducted by a member of the Machine Learning in Reserving Working Party. It should be of interest to actuaries at all levels in ML, especially those keen on the practical implementation of ML in the reserving process. Attendees will have access to an accompanying Python notebook with full workings for further reading, enabling them to fully replicate the models.