基于数学统计的保险赔付风险预测模型设计

Design of insurance claim risk prediction model based on mathematical statistics

设计基于数学统计的保险赔付风险预测模型,分别从使用量、驾驶表现、危险驾驶、出行习惯四个方面选取能够反映驾驶行为的20个风险因子构建指标体系,利用数学统计中的因子分析法从上述指标体系内选取6个能代表驾驶行为风险情况的典型风险因子;以选取的典型风险因子为基础结合二分类随机变量,利用具有优秀分类与回归性能的XGBoost模型构建保险赔付风险预测模型,预测变量所属类别与概率分布。实证分析结果显示,该模型迭代速度较快,AUC值与F值相较于传统Logistic模型分别上升67.4%和2.3%,显著高于对比模型。

A risk prediction model for the insurance compensation is designed on the basis of mathematical statistics,and20 risk factors that can reflect driving behavior are selected in four aspects of usage amount,driving expression,dangerous driving and traveling habit to construct the index system.Six typical risk factors that can represent the risk situation of driving behavior are selected from the above index system by means of the factor analysis method in mathematical statistics.On the basis of typical risk factors,the XGBoost model with excellent classification and regression performance is used to build the insurance claim risk prediction model in combination with two dichotomy random variables for the prediction of the variables′category and probability distribution.The results of empirical analysis show that the iteration speed of the model is faster,and AUC value and F value are increased by 67.4%and 2.3%respectively in comparison with the traditional Logistic model,which are significantly higher than those of the compared model.