基于随机森林算法的云南昆磨高速公路气象风险研究

Meteorological Risk Research of Kunmo Expressway in Yunnan Province Based on Random Forest Algorithm

利用2018—2021年昆磨高速沿线交通气象观测站点数据,计算了总降水量及能见度低于500 m的天数,将其与地灾点、隧道点和桥梁点核密度以及道路曲率半径栅格数据输入到基于随机森林算法的预测模型中,最终昆磨高速危险路段回归预测结果R 2值为0.790,P值为0.001,满足显著性检验要求,且预测结果与验证数据之间高度拟合。结果表明:①中度危险性以上路段集中在昆磨高速沿线“上—中”段的呈贡区、晋宁县、红塔区、峨山县、宁洱县以及景洪市,其中重大危险等级路段主要分布在景洪市和宁洱县;②从随机森林算法预测结果来看,隧道点核密度重要性占比为26%,能见度小于500 m的天数重要性占比为24%,2项综合占比为50%,说明隧道的分布和能见度低的天气状况对昆磨高速沿线行车安全影响最大。

Based on the data of 22 traffic meteorological observation stations along Kunming-Mohan Expressway from 2018 to 2021,the total precipitation and the number of days with visibility less than 500 meters are calculated.The kernel density of ground disaster points,tunnel points and bridge points and the grid data of road curvature radius are input into the prediction model based on random forest algorithm.Finally,the R 2 value of the regression prediction result of the dangerous section of Kunming-Mohan Expressway is 0.790,and the P value is 0.001,which meets the requirements of significance test,and the prediction result is highly fitted with the verification data.The results show that:①The sections above moderate risk are concentrated in Chenggong,Jinning,Hongta,Eshan,Ning er and Jinghong City along the upper-middle section of Kunmo Expressway,among which the sections with major risk levels are mainly distributed in Jinghong City and Ning er County;②According to the prediction results of random forest algorithm,the importance of tunnel point nuclear density accounts for 26%,the importance of days with visibility less than 500 meters accounts for 24%,and the two comprehensively accounts for 50%,indicating that the distribution of tunnels and the weather conditions with low visibility have the greatest impact on the driving safety along Kunmo Expressway.