变点问题统计分析框架及应用——以我国财险赔付支出数据为例

Statistical Analysis Framework and Application of Change Point Problem:Taking China's Property Insurance Payment Data as an Example

经济金融领域的数据通常具有结构性变化特征,其突变点往往不易辨别,因此对关键风险点的识别和差异化建模至关重要。在已有研究基础上,由变点识别、检验、估计为主线构建了变点问题研究的分析框架,并以保监会每月发布的赔付支出数据为例识别了一年中宏观财险赔款的周期性突变点,根据变点前后差异建立了不同的损失风险模型。研究发现:具有单一变点下的数据服从"双峰"分布,当系统内存在变点时,混合模型优于单一模型。

The data in the economic and financial field are usually characterized by structural changes,and their mutation points are often difficult to identify.Therefore,it is very important to identify and differentiate the key risk points.On the basis of the existing research,the analysis framework of change point problem is constructed with the main line of change point identification,test and estimation.Taking the monthly payment data issued by the CIRC as an example,the periodic change point of the macro-financial insurance payment in one year is identified,and the different loss risk models are built according to the difference before and after the change point.It is found that the data with a single change point obeys the“Double Peak”.When there is a change point in the system,the mixed model is better than the single model.