Automatic Personalized Health Insurance Recommendation Based on Utility and User Feedback

Health insurance is a fundamental type of insurance that is essential to every person. Due to the increasing requests for personalized health insurance recommendation, traditional ways of obtaining recommended health insurance products from experts or agents are insufficient, and thus automatic personalized health insurance recommendation is needed. However, existing approaches to addressing this problem either need heavy input from users to obtain the utility of health insurance products for recommendation, or directly rely on user feedback and neglect the utility in recommendation. In this paper, we propose an automatic personalized health insurance recommendation system that incorporates both utility-based recommendation and user feedback based recommendation. To achieve that, we propose a two-component architecture in our system, where the first component is for daily operations to provide utility-based personalized recommendation for users with simple input of health risks, and the second component is to refine the first component by learning dynamic user preferences from user feedback data and combine the learnt preferences into utility-based recommendation. On a complex real-world insurance dataset, our proposed system achieves 2x more effectiveness than the baseline approaches for improving the recommendation quality.