摘要
为探究气象因素对交通事故的影响以给保险公司的车险业务提供科学的信息支持,本文选取北京、天津、大连、青岛、上海、重庆、深圳、厦门、宁波九个城市作为研究区域,并利用其2013-2017年的车辆出险次数数据,以及对应气象站点的观测资料,详细探讨了气象条件对车辆出险的影响,基于Prophet时间序列模型,构建了混合回归模型,并对出险次数进行预测。结果表明:气象条件中对出险次数影响较大的几个因素为能见度、温度、相对湿度、降水量,且与气象要素的等级有关。24小时降水量50mm以下时平均日出险次数最高。日最高温度超过32℃后,平均出险次数升高;发生在0℃的车辆出险次数低于0℃以上的。相对湿度升高时,日出险次数降低。10.7m/s的极大风速与平均出险次数无明显相关关系;极大风速7m/s以下时,出险次数与风速呈负相关关系。能见度高于1km时,对出险次数无明显影响;1km以下,能见度越低,出险次数越高。以时间序列预测值作为间接变量并添加气象变量,按照逐步回归的方式选择变量,构建的混合回归模型拟合优度比只有时序模型高。
In order to explore the impact of meteorological factors on traffic accidents and provide scientific information to the insurance company's auto insurance business,this paper selects nine cities of Beijing,Tianjin,Dalian,Qingdao,Shanghai,Chongqing,Shenzhen,Xiamen and Ningbo as research areas,and uses the data of automobile accidents from 2013 to 2017 and the observed meteorological data.The influence of meteorological conditions on vehicle risk is discussed in detail,and a mixed regression model is constructed,and the number of accidents is predicted.The research results are as follows:Several factors that have a greater impact on the number of out-of-risk conditions in meteorological conditions are visibility,temperature,relative humidity,and precipitation,and are related to the level of meteorological elements.The average number of daily motor vehicles accidents is the highest when the 24-hour precipitation is below 50mm.After the daily maximum temperature exceeds 32 ℃,the average number of accidents increases;the number of vehicles that occur below 0 ℃ is less than that above 0 ℃.When the relative humidity increases,the number of accidents decreases.There is no significant correlation between the maximum wind speed over 10.7m/s(that is,the wind over the strong wind)and the average number of accidents;when the wind speed is below 7m/s,the number of accidents is negatively correlated with the wind speed.When the visibility is higher than 1km,there is no obvious influence on the number of dangers;below 1km,the lower the visibility,the higher the number of risks.Using the predicted value of the Prophet model as an indirect variable and adding meteorological variables,the variables are selected according to the stepwise regression method.The constructed goodness of the mixed regression model is higher than that of the time series model.
出处
《保险职业学院学报》
2019年第1期5-11,共7页
Journal of Insurance Professional College
