基于径向基神经网络的高速公路交通事故影响因素分析

Analysis of Influencing Factors of Highway Traffic Accidents Based on Radial Basis Neural Network

为实现多重因素作用下的高速公路交通事故分析,基于径向基神经网络理论建立交通事故预测模型,综合考虑道路线形(圆曲线半径、转角、纵坡)、交通组成(自然交通量、折算交通量、货车率)及路面技术状况指标(PCI、RQI、RDI、SRI)三类因素,对某高速公路2018年—2022年的交通事故进行建模训练,并分析各因素对诱发交通事故的贡献。结果表明:径向基神经网络用于预测交通事故,可行性强、准确率高(91.6%);相较于线形指标而言,交通组成与路况指标对交通事故影响较大;各因素对诱发交通事故的贡献排序为自然交通量>SRI>折算交通量>RDI>货车率。

To analyze highway traffic accidents under the influence of multiple factors,a traffic accident prediction model based on the radial basis neural network theory was established.This model considered three types of factors regarding road alignment(radius of circular curve,turning angle,longitudinal slope),traffic composition(natural traffic volume,converted traffic volume,truck rate),and pavement technical condition index(PCI,RQI,RDI,SRI).The prediction model was trained using the traffic accident data from 2018 to2022,and the contributions of each factor to induced traffic accidents was analyzed.The results show that the radial basis neural network can be used to predict traffic accidents with high feasibility and accuracy(91.6%).The traffic composition and road condition indicators have the greater influence on traffic accidents than the alignment indicators.The ranking of factors contributing to traffic accidents is as follows:natural traffic volume SRI converted traffic volume RDI truck rate.