《保险研究》20200905-《考虑老年痴呆症的医疗险住院费用预测与比较——基于机器学习模型》（赵颖旭、包竹青、高珊、周亮、刘逸圣、陈浩、张宁）

［摘   要］老年痴呆症严重影响老年人生活质量,其影响程度也随着中国老年人口比例提升和预期寿命增加而不断加深。在此背景下,针对老年痴呆症的商业健康险开发意义重大。本文基于详实的历史数据,利用传统定价中估计定价因素的广义线性模型对老年痴呆症患者住院费用进行估算和预测,同时用随机森林、LASSO回归、LightGBM三种机器学习方法进行同样的估算预测,以期为保险公司开发产品、丰富保险市场、助力养老提供基础。对四种方法进行横向比较的结果显示,机器学习方法除了在结果估算方面具有更大的优势外,还在影响因素的数量、精准定价以及模型适应方面具有更大的潜力。最终的计算结果和模型使用的过程方法结合住院率数据可用于具体测算老年痴呆症医疗保险费用。

［关键词］老年痴呆症；商业医疗险；带病投保；机器学习；保费测算

［基金项目］本文得到中央高校基本科研业务费专项基金和教育部人文社会科学重点研究基地重大项目（16JJD790060）资助。

［作者简介］赵颖旭,泰康健康产业投资控股有限公司首席医疗分析师,流行病学博士,中央财经大学中国金融科技研究中心兼职研究员；包竹青,泰康保险集团股份有限公司运营企划主管；高珊,泰康保险集团股份有限公司运营企划专员；周亮,中央财经大学中国金融科技研究中心兼职研究员；刘逸圣,泰康健康产业投资控股有限公司实习生；陈浩,中央财经大学保险学院研究生；张宁(通信作者),中央财经大学金融学院教授,中央财经大学中国金融科技研究中心主任。

Projection and Comparison of the Hospitalization Expenses of Medical Insurance Covering Alzheimer′s Disease—Based on the machine learning model

ZHAO Ying-xu,BAO Zhu-qing,GAO Shan,ZHOU Liang,LIU Yi-sheng,CHEN Hao,ZHANG Ning

Abstract：Alzheimer′s disease seriously affects the quality of life of the elderly,and its impacts continue to deepen as the proportion of the aged keeps rising and the life expectancy increases in China.The development of commercial health insurance for Alzheimer′s disease becomes more and more important.Based on detailed historical data,the paper used the generalized linear model to estimate pricing factors to predict the hospitalization costs of Alzheimer′s patients.Meanwhile,it also applied the machine learning methods like random forest,LASSO regression and LightGBM to estimate and predict hospitalization costs of Alzheimer′s patients.This research is of great significance for insurance companies to develop diversified products,enrich the insurance market,and help the elderly.The horizontal comparison of the four methods show that machine learning methods have greater advantages in cost estimation,and also have greater potential in the number of influencing factors,precise pricing and model adaptation.The final calculation results and the process method combined with the hospitalization rate data can be used to specifically calculate the medical insurance cost of Alzheimer′s disease.

Key words：Alzheimer′s disease;commercial medical insurance;applying for insurance with existing disease;machine learning;premium calculation