《保险研究》20241205-《基于组合机器学习模型的我国长期护理保险产品定价策略》（程恭品、沈世杰、徐冬妮）

［摘   要］大数据时代，机器学习技术在预测精度和计算效率上展现出独特优势，将传统的统计分析模型与机器学习算法相结合，构建更为精确的我国长期护理保险产品定价方法成为一种积极有益的探索。本文依据国际失能标准，将老年人健康状态划分为六种，利用CHARLS数据库2018年和2020年的数据，将被保险对象扩展至40岁以上，采用基于PSO算法的XGBoost-Logistic组合模型分析健康状态影响因素，使用双向长短时记忆网络(BiLSTM)测算转移概率，使用CBD-LSTM组合模型测算预期寿命，从而对我国长期护理保险产品进行精准定价。

［关键词］长期护理保险；XGBoost-Logistic组合模型；BiLSTM模型；CBD-LSTM组合模型

［基金项目］本文受到国家自然科学基金青年项目(11801265、11501211)资助。

［作者简介］程恭品(通讯作者)，南京财经大学经济学院副教授、硕士生导师，研究方向：金融统计，保险精算，随机决策优化；沈世杰，南京财经大学硕士研究生；徐冬妮，南京财经大学硕士研究生。

Pricing Strategy of Long-term Care Insurance Products in China Based on Combinatorial Machine Learning Model

CHENG Gong-pin,SHEN Shi-jie,XU Dong-ni

Abstract：In the era of big data,machine learning technology has shown unique advantages in prediction accuracy and computational efficiency.Combining traditional statistical analysis models with machine learning algorithms to construct more accurate pricing methods for long-term care insurance products in China has become a positive and beneficial exploration.This article divides the health status of elderly people into six categories based on international disability standards.Using data from the CHARLS database in 2018 and 2020,the insured population is expanded to be over 40 years old.The XGBoost Logistic combination model based on PSO algorithm is used to analyze the factors affecting health status.The bidirectional long short-term memory network (BiLSTM) is used to calculate the transition probability,and the CBD-LSTM combination model is used to calculate the expected life expectancy,in order to accurately price long-term care insurance in China.

Key words：long term care insurance;XGBoost-Logistic combination model;BiLSTM model;CBD-LSTM combination model