KDD For Insurance Risk Assessment: A Case Study

Insurance is a business of risks. Identifying and understanding areas of risk is an important task performed by an insurer. An assessment of risk is used to set the appropriate premium for insurance policies. This paper describes a KDD exercise which uses decision tree techniques to identify significant areas of risk within an insurance portfolio. The real world dataset used contains information about policies and insurance claims on those policies. Decision trees can be constructed to identify and describe areas of high risk which are then evaluated, as a separate exercise, in terms of claim frequency and claim costs. The paper stresses the idea of interactive post-processing, or evaluation, of the patterns that are illuminated by traditional data mining tools.