How Unlucky People Continue to be Unlucky: : A Study of the Predictive Capabilities of Insurance Claim Data

Abstract: This bachelor thesis in the field of mathematical statistics was carried out in collaboration with an upcoming insurance start-up, Hedvig, and had the objective of investigating the predictive capabilities of different types of insurance claims. This was done through regression analysis, and more specifically the area in regression analysis called generalized linear models. Logistic regression was employed as the modeling technique, and data points were modeled in various ways to then be used to fit models in order to determine the most optimal one based on some pre-determined statistical evaluation metrics. The final model had an accuracy of above 96%, and the results showed that certain types of claims had a bigger contribution to the probability of a claim occurring the next period. This study contributes to the understanding of the predictive capabilities of insurance claim data and provides insights that could aid in the development of more accurate and efficient insurance pricing models.