最大熵方法下的信度估计

The Maximum Entropy Method to the Credibility Estimation

  • 摘要: 在经典的信度理论中, 计算信度保费时除了要知道历史索赔数据, 还要对的分布做出一定假设. 本文在索赔数据的分布信息完全未知时, 运用最大熵的方法给出了相应的信度保费. 并且在一定的假设条件下得到两个性质, 即: 对年份较近的索赔比年份较远的索赔赋予更大的权重; 经典的信度保费是本文信度保费的一种特殊情况. 最后进行数值模拟, 结果表明: 在均方误差标准下, 本文的结果优于其它三个模型.

     

    Abstract: Besides the claims data in the past, certain assumptions about the distribution of claimsare required to derive the credibility premium in the classical theory. In the paper, the credibility premium can be calculated via the maximum entropy method if we know nothing about the distribution of claims . Furthermore, two corollaries are obtained under certain assumptions, that is, new claims have more weight than the old ones and the classical credibility formula is a special case of the credibility premium derived in the present paper. Finally, the simulation study is presented to illustrate that the credibility premium in the present paper is better than other models if the mean square error is taken as the evaluation criterion.

     

/

返回文章
返回