City Research Online

Abstract

Using a modern Bayesian implementation technique, this thesis shows two applications of individual modelling in real data sets. The simulation approach is adopted, with a Markov chain Monte Carlo (MCMC) method - Reversible Jump MCMC - as the core of the thesis. This technique allows the definition of a model with few underlying assumptions and based on a changing-dimension parameter set.

Its first application is in automobile insurance, where the model estimates at the same time the number of groups and their respective risk parameters in order to have a better description of the analysed data. Since all this process is based on a continuous piecewise distribution, no obvious analytical solution for this type of problem is available. RJMCMC is the only stochastic simulation that allows this change of dimensionality.

The flexibility of this model is explored in the second application presented in this thesis. In this new case, the aim is not to define the number of groups, but to use a limited number assumptions to model the claim reserves in a dental insurance coverage.

Both applications model frequency and severity separately and apply the grouping technique to both discrete and continuous variables.

Publication Type:	Thesis (Doctoral)
Subjects:	H Social Sciences > HG Finance Q Science > QA Mathematics
Departments:	School of Science & Technology > Mathematics

Export

Downloads