- Posthuma Partners
- Actuary, Consultancy, Data Science, IT
- Required language
- Dutch, English
Interested in writing your thesis at Posthuma Partners? Hereby a number of possibilities
One of the innovative products of Posthuma Partners is CMF, the Claims Management Filter. CMF is an AI-enabled tool for the classification of insurance claims. Based on the claim amount, claims are classified as “expensive”, “moderate” or “inexpensive”. We see two interesting subjects around CMF which lend themselves for thesis-research.
The classification is done by the quantile corresponding to the claim amount. The underlying model that produces these quantiles is currently a heteroscedastic regression model. You would investigate how a heteroscedastic regression model compares to alternative models, e.g., neural networks or quantile regression models. Questions you would look into are: what is the relative performance of the alternatives, are there specific advantages or pitfalls, how to explain the models to non-experts? A nice project to apply lesser-known methods to real-world data-sets.
Another line of investigation is to look at ways to select a-typical claims. Can this be done by appropriate cluster or tree-based methods? Do different methods select different claims? How to explain that a selected claim is a-typical? Is there business-value in identifying these claims? When it comes to insurance claims, this area is virtually unexplored and requires imagination and creativity.