【摘要】 随着人们健康意识不断增强，医疗保障需求呈现多元化特点。为进一步推动商业医疗保险健康可持续发展，对商业医疗险赔付的精准预测研究很有必要。基于包括个人行为偏好的经验数据，构建并对比商业医疗险的创新预测机器学习模型，分析与医疗险赔付风险相关的重要因素，研究发现，个人对健康信息的关注等行为因素与赔付风险高度相关，可以为年龄、性别、受教育水平、婚姻状况、地区等传统赔付经验分析因素提供很好的补充。与索赔发生关联最大的是年龄，与案均医疗赔款关联最大的是赔付次数，远高于其他因子。当赔付次数多于5次时，预测案均医疗赔款呈发散性分布。被保险人为女性、学历更低的人更易发生索赔。疾病负担呈现区域不平衡特点。据此提出完善商业医疗险风险管理等建议。更多还原

【Abstract】 As people’s health awareness continues to grow, the demand for medical insurance is characterized by diversification. To further promote the healthy and sustainable development of commercial medical insurance, it is necessary to study the accurate prediction of commercial medical insurance payouts. Based on empirical data including individual behavioral preferences, innovative predictive machine learning models for commercial medical insurance are constructed and compared to analyze the important factors associated with the risk of medical insurance payouts. It is found that behavioral factors such as individual’s attention to health information are highly correlated with claim risk, which can provide a good complement to traditional factors of claim for empirical analysis such as age, gender, education level, marital status, and region. Age is the most associated with claim risk, and the times of claims is the most associated with claim amount, which are much more influential than other factors. When someone claims insurance for five times or more, the predicted average medical claims per case presents a dispersed distribution. Claims are more likely to occur when the insured is a woman or has a lower level of education. The disease burden is characterized by regional imbalance. Accordingly, recommendations are proposed to improve the risk management of commercial medical insurance.更多还原