Skip to main content
To KTH's start page To KTH's start page

Dennis Uygur Andersson: Optimizing method selection for IBNR-reserve calculation using machine learning

Master thesis final presentation

Time: Wed 2023-06-07 10.40

Location: Meeting room 9, floor 2, house 1, Albano

Respondent: Dennis Uygur Andersson

Supervisor: Filip Lindskog

Export to calendar


This work implements an approach which were introduced by Caesar Balona & Ronald Richman in their article The Actuary and IBNR Techniques: A Machine Learning Approach (2021), that combines the strengths of both traditional and machine learning reserving methods. This approach is still based on the ordinary reserving methods available today, such as chain ladder and Bornhuetter–Ferguson method, with the modification that we vary how the loss development factors are estimated and included/excluded. This was achieved using AvE and CDR as score tests. The estimated reserves using the machine learning approach were then compared to corresponding reserves using the customary methods. The outcome showed that the ordinary reserving methods, especially the chain ladder method, overall performed better, even though we sometimes gained better results using the new approach. Finally, we discuss how this thesis could be improved until next time.