基于MCMC模拟的索赔次数模型

Applications of MCMC Simulation to Claim Counts Models

对非寿险保险公司来说,精确预测索赔次数是保险费率厘定中的一个重要环节。传统上对索赔次数模型的估计是基于极大似然估计方法,模型参数是从数据中估计而来。但是由于造成索赔次数的原因很多,索赔次数服从的具体分布通常很复杂,仅用已知数据估计模型参数存在很大偏差。本文从另一角度,基于MCMC方法估计索赔次数模型。研究表明,利用MCMC模拟可以动态模拟出未知参数的后验分布,从而提高索赔次数模型的估计精度。

Precise prediction of claims is important in insurance ratemaking in non--life insurance companies. In tradition, claim counts model is estimated by maximum likelihood estimation method, so parameters are estimated from data. But there are many reasons inducing claim counts. And distribu- tions of claim counts are usually complex. So this leads to estimators of parameters that are often biased. In this paper, we apply MCMC simulation to estimate parameters of claim counts models. It indi- cates that MCMC method can dynamically simulate parameter posterior distribution and improve estima-tion precision.