基于高速公路事故数据的自适应事故持续时间预测模型研究

打开文本图片集

中图分类号：U491.31 文献标志码：A文章编号：1002-4026（2025）05-0104-11

Abstract：Freewaytraffc accidents seriously affect road safetyandaccessbilityAccurately predicting the durations of accidents is keytoimproving emergencyresponse eficiency，alleviating traffccongestion，and reducingthe risk of secondaryaccidents.This paper proposes anadaptive parameter-optimization model based ona deep belief network （DBN）andgenetic algorithm（GA）for predictingtraffcaccident durations.Trafficaccident data from freeways in Shandong province were colected from 2020to2022，including16 variables suchasroad，temporal atributes，and environmentalatributes.The Spearmancorrelation coeffcientandboxplots were usedtoanalyzethecorrelation between each variable and the accident duration，ensuring the validityand significance of the selected variables.Based on this analysis，we developedanadaptive parameteroptimization-based prediction model，GADBN，usingnumerous traffic accidentdata.Thismodelintegratestheglobalsearchandoptimizationcapabitiesof theGAtonotablyimprovethe predictive accuracyof the DBN.To validate the model effctiveness，experimental comparisons were conducted with other algorithms such assupport vector regression，radialbasis functions，XGBoost，and DBNs，with mean absolute percentage error （ δMAPE ） and root-mean-square error （ δRMSE ）being used as evaluation metrics. The experimental results showed that the GADBN model achieved δΠMAPE and values of 16.49% and 9.12，respectively，outperforming the other comparison models，thereby demonstrating its effectiveness and practicality.

Key words ： traffic safety； traffic accidents；accident duration；adaptive parameter optimization； freeway

截至2022 年底，我国高速公路总里程已达到17.7万公里，庞大的基础设施在带来便利的同时，也带来了严重的交通安全问题[1]。（剩余12208字）