The Application of Crisp and Fuzzy Decision Trees to Monitor Insurance Customer Database

Insurance has become an important way of investment, savings and risk management. Abnormal accessing of customer data has been recognized as a big cause to loss of insurance companies' benefits and reputation. In this paper, we analyze the behavior of abnormal accessing from the operational records of operators, and then generate the crisp and fuzzy decision trees to classify the categories of operators' behavior, so the report of abnormal accessing can be conducted in the follow steps. This offers some support for the related management of abnormal accessing database. After testing experiments, the fuzzy decision tree is more effective than the crisp approach.