THE INFLUENCE OF UTILITY FUNCTIONS ON INSURANCE CHOICES

Abstract

As society advances, people's quality of life is improving. More and more people are not only paying attention to the physical quality of life but also focusing on the wealth management such as buying insurance to support their life quality. Because of this, insurance companies are providing a growing number of policies to satisfy the public need. Utility is a measurement of people's welfare. We aim to estimate a utility function model based on the people's preferences. However, the following reasons make the utility function hard to estimate: First of all, for existing models, it is difficult to apply these models to describe multifarious people's preferences. Secondly, it is difficult to obtain the exact utility model from datasets. The primary issue is how to establish a good predictive model for all different insurance needs. In this thesis, we are trying to develop a framework for selecting suitable insurance choices. We compute the amount of insurance to purchase for a range of randomly generated utility functions and a range of situations. We then apply data mining approaches, such as Principal Component Analysis (PCA) and Singular Value Decomposition (SVD) to determine which aspects of the utility function are most in uential on insurance choice, and use these aspects to reduce the dimension of the predictive model. Finally, we test the reduced dimension of predictive models on the simulated datasets. Our results demonstrate that the some reduced predictive models can get high predictive accuracy under a range of conditions.