Artificial Intelligence in Insurance Profitability Models

Artificial intelligence techniques have been successfully applied to the underwriting of credit risk products such as credit cards and loans, but not to the underwriting of casualty risk. Techniques in data mining and neural networks provide an efficient and effective method for creating a decision-aiding tool for underwriters. A web-based back propagation program will increase the accessibility of the tool, which outperforms traditional statistical techniques such as multiple linear regression. The application will be implemented on publicly available German credit data, illustrating the benefits of neural networks in information technology. Thesis Supervisor: David R. Karger