Industrial risk classification using credibility theory and hierarchical clustering

Industrial risk classification is essential for Workers Compensation Boards to review and update their insurance policies for each fiscal year. This thesis develops two statistical approaches, namely the implementation of credibility theory and a novel hierarchical clustering method, to perform classification based on workplace risk levels for different industries. Real data that contains the historical claim experiences as well as payroll amounts are used to illustrate the methods. We also compare our results with the grouping result obtained by a consulting firm to check the validity of the proposed methods. The reasons of discrepancy, if any, are listed for reference and future improvement.