Clustering and Behavioral Analysis of Health Insurance Owners

For an insurance company, accurately analyzing its customers and being able to differentiate them descriptively and behaviorally are processes that affect premium calculation and indirectly the company's revenue. Thanks to increasing data volumes, these processes are handled with the use of advanced analytical methods rather than expert interpretationbased methods. In this study, k-means clustering models are trained using descriptive features such as gender, age, province and behavioral features such as recency, frequency and diversity, which are not commonly used in similar studies. With the decision trees trained using the groups assigned by the models as target variables, rule extraction was performed to ensure the interpretability and explainability of the models for the end user and the financial risks of each group for the company were calculated.