
General Insurance Spring Webinar Series 2026
The Machine Learning in Reserving Working Party shares its recent work on our Neural Network Framework for reserving on individual claims data.
This presentation covers:
- An overview of the Neural Network Framework
- How the structure of the data affects model performance
- A comparison of a neural network with the chain ladder
- Findings on time series based neural networks, where a GRU (gated recurrent unit) performs notably better than a basic neural network on reserving data.
All the code behind this research is publicly available, and we encourage reserving actuaries to try these methods on their own data.
This talk is supported by the following blog posts which include links to downloadable notebooks:
If you are new to neural networks, we recommend starting with our 2024 Spring Conference talk and its accompanying blog series.