基于DNN的有限时间期望折现罚函数研究
首发时间:2023-03-15
摘要:本文主要研究了基于常数利息力投资的复合泊松风险模型中有限时间期望折现罚函数的快速计算。在逼近此类有限时间期望折现罚函数时,本文采用以数据驱动为基础的深度神经网络(DNN)方法,同时结合该模型下有限时间期望折现罚函数满足的一般偏积分微分方程(PIDE),详细讨论了其求解过程。此外,本文还给出了该模型下有限时间期望折现罚函数处于不同情况时的数值算例,并对不同网络参数设置、不同索赔分布的结果分别进行对比,最终得出相关结论。
关键词: 破产概率; 有限时间期望折现罚函数; 偏积分微分方程; 利息力
For information in English, please click here
Research on the finite-time expected discounted penalty function based on DNN
Abstract:This paper focuses on the fast calculation of the finite-time expected discounted penalty function under the compound Poisson model with consideration of a constant interest force investment. When approaching this kind of finite-time expected discounted penalty function, this paper uses the data-driven deep neural network method(DNN), and discusses its solution procedure in detail in combination with the general partial integral differential equation (PIDE) which is satisfied by the finite-time expected discounted penalty function under this model. In addition, this paper also gives some numerical examples of this type of the finite-time expected discounted penalty function under different conditions, and compares the numerical results of different network parameter settings and different claim distributions. Finally, this papre draws some relevant conclusions.
Keywords: Ruin probability Finite-time expected discount penalty function PIDE Interest force
基金:
引用
No.****
同行评议
勘误表
基于DNN的有限时间期望折现罚函数研究
评论
全部评论