University of Twente Student Theses


Adapting a Hierarchical Gaussian Process model to predict the loss reserve of a non-life insurer

Ruitenberg, P.L. (2019) Adapting a Hierarchical Gaussian Process model to predict the loss reserve of a non-life insurer.

[img] PDF
Abstract:Potential improvements are researched of a hierarchical Gaussian process model on the actuarial challenge of predicting the Loss Reserve of a non-life insurer. The model performance and design is validated, by assessing the model performance on a more extensive data set. Furthermore, design choices will be validated by varying the prior distributions on hyperparameters in the model to other weakly informative priors used in literature on Gaussian processes. Also, extending the model with external information has been researched. Various methods are researched to incorporate premium information in the model. The results indicate that the performance on volatile run-off triangles by the model of Lally and Hartman (2018) still has room for improvement. The design choices of the prior distributions of the hyperparameters originally made are adequate. For the model in general, we conclude that most weakly informative priors give adequate results. A transformation to Loss Ratio’s has a positive effect on the prediction of the Best Estimate of the Loss Reserve. Adding a Bornheutter-Ferguson estimation to the model gives good results regarding to the uncertainty of the prediction. However, this implementation is reliant on the quality of these estimations. We recommend testing the model performance on incremental data, or run-off triangles of incurred claims. We also recommend tuning the kernel function of the model to triangle specific characteristics. As for implementing external data in the model, other data such as the number of claims could be researched. Moreover, validating a faster Gaussian process approximation is recommended.
Item Type:Essay (Master)
KPMG Advisory N.V., Amstelveen, The Netherlands
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:83 economics
Programme:Industrial Engineering and Management MSc (60029)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page