保险科技影响下的基于贝叶斯网络的保费定价研究

With the development of insurance technology, the gaining information and deep mining abilities of Internet participants have largely improved. Therefore, the actuarial area is also supposed to develop a new pricing model to adapt these changes. In classical actuarial models, the premium price depends on historical claim data and the premium system is static homogenization, but such system may cause serious adverse selection risk and fraud risk. Based on the core issue of actuarial science: premium pricing, this paper further refines and subdivides the risk factors and builds a dynamically adjusted premium system to reduce the degree of asymmetric information. The development of Internet and big data technology makes the Bayesian network (BN), a method to process uncertain and incomplete information, to apply into the pricing of Internet insurance. This study replaces tradition missing data with network data, and uses the combination of the priori of cluster analysis and posteriori claim data instead of traditional actuarial pricing models, then builds a new pricing system to adapt current development of insurance technology. Meanwhile, considering the problem of insurance risk measurement, ruin probability, a crucial actuarial risk, is used as the risk-pre-warning indicator, and the dynamic premium system is solved by dynamic stochastic programming, then, the ruin probability estimation method of dynamic premium system can be obtained at the same time, which develop the actuarial pricing theory. Finally, based on the results, warranty designs and policy advice for Internet insurance companies are proposed.

随着保险科技的发展，互联网参与方获取和深度挖掘信息的能力大幅提高，保险精算领域也应该发展出新的定价模式来适应这种变化。在经典的精算模型中，保费价格依赖于历史索赔数据且是静态同质化的，这种保费系统会带来严重的逆向选择风险和欺诈风险。本项目从精算保险的核心问题－保费定价入手，对风险因子进一步地提炼和细分，建立动态调整的保费系统，力图降低信息不对称程度。互联网以及大数据技术的发展使得贝叶斯网络这一处理不确定，不完全信息推理的方法应用到互联网保险的定价上。用网络数据替代传统损失数据，应用聚类分析的先验与后验索赔数据结合的方法替代传统的精算定价模型，建立适用于当今保险科技发展的定价系统。同时考虑保险风险测算问题，用破产概率这一重要的精算风险指标作为风险预警的指标，采用动态随机规划方法求解动态保费系统，同时可以得到动态保费系统的破产概率的估计方法，从而发展了精算定价理论。最后根据理论结果提出政策建议。

随着保险科技的发展，互联网参与方获取和深度挖掘信息的能力大幅提高，保险精算领域.也应该发展出新的定价模式来适应这种变化。在经典的精算模型中，保费价格依赖于历史索赔.数据且是静态同质化的，这种保费系统会带来严重的逆向选择风险和欺诈风险。本项目从精算.保险的核心问题－保费定价入手，对风险因子进一步地提炼和细分，建立动态调整的保费系统.，力图降低信息不对称程度。互联网以及大数据技术的发展使得贝叶斯网络这一处理不确定，.不完全信息推理的方法应用到互联网保险的定价上。用网络数据替代传统损失数据，应用聚类.分析的先验与后验索赔数据结合的方法替代传统的精算定价模型，建立适用于当今保险科技发.展的定价系统。同时考虑保险风险测算问题，项目组从保险基金动力系统的稳定性展开研究，.测算了带有脉冲冲击的保险基金动力稳定性的条件，从而发展了精算定价理论。最后根据理论结果提出政策建议。