An approach to model complex high-dimensional insurance data

This paper describes common features in data sets from motor vehicle insurance companies and proposes a general approach which exploits knowledge of such features in order to model high-dimensional data sets with a complex dependency structure. The results of the approach can be a basis to develop insurance tariffs. The approach is applied to a collection of data sets from several motor vehicle insurance companies. As an example, we use a nonparametric approach based on a combination of two methods from modern statistical machine learning, i.e. kernel logistic regression and ε−support vector regression.