Master's Thesis presentation, Oskar Molin and Viktor Lidman

This thesis is centered on creating a geographical risk model for pricing of property content insurance in Sweden. As insurance is a fundamental financial service, correct and fair pricing is important from both a business and societal point of view. In this thesis we try to improve upon the price model by modeling geographical risk. The project compares the classical approach of Generalized Linear Models (GLMs) with novel Gradient Boosting Machines (GBMs) and a spatial epidemiological method called the Besag-York-Mollié (BYM) model. The results show that GBMs can outperform GLMs and the methods we propose can lead to improved insurance price models.