基于改进深度森林算法的高速公路交通事故风险预测  

Highway traffic accident risk prediction based on improved deep forest algorithm

作　　者：张浩[1] ZHANG Hao(Faculty of Engineering,China University of Geoscience(s Wuhan),Wuhan 430074,China)

机构地区：[1]中国地质大学(武汉)工程学院,湖北武汉430074

出　　处：《安全与环境工程》2024年第6期91-99,共9页Safety and Environmental Engineering

基　　金：湖北省安全生产专项资金科技项目(SJZX20230904)。

摘　　要：高速公路交通事故风险预测对于实行动态交通安全管理至关重要。为探究影响高速公路交通事故风险的主要因素以及准确预测高速公路交通事故风险,提出了一种基于改进深度森林算法的高速公路交通事故风险预测模型。首先以高速公路交通事故数据、交通流数据、天气数据、道路条件和特殊时间段数据为基础,选取了能够表征高速公路交通事故风险的特征变量,并采用随机森林算法对特征变量的重要度进行了计算,筛选出对高速公路交通事故风险影响较大的重要特征变量,以解决后面计算过程中的维度灾难问题;然后运用基于决策树的LightGBM和XGBoost算法对深度森林模型的级联森林结构进行了改进;最后将改进深度森林算法应用于高速公路事故风险预测。结果表明:与现有的SVM、随机森林和深度森林算法相比,改进深度森林算法具有更优的预测性能,其预测准确率达到了88.84%,预测结果能为高速公路交通管理部门制定更为有效的安全管控措施提供决策支持。Highway traffic accident risk prediction is very important for dynamic traffic safety management.In order to explore the main factors affecting the highway traffic accident risk and accurately predict the highway traffic accident risk,a highway traffic accident risk prediction model based on improved deep forest algorithm is proposed.Firstly,based on expressway traffic accident data,traffic flow data,weather data,road conditions and special time period data,the characteristic variables that can represent highway traffic accident risk were selected.The random forest algorithm was used to calculate the importance of the characteristic variables,and screen out the important characteristic variables that have a greater impact on the highway traffic accident risk,and solve the dimensional disaster problem in the following calculation process.Then the cascade forest structure of the deep forest model was improved by using LightGBM and XGBoost algorithms based on decision tree.Finally,the improved deep forest algorithm was applied to highway accident risk prediction.The results show that compared with the existing SVM,random forest and deep forest algorithms,the improved deep forest algorithm has better prediction performance,and the accuracy rate reaches 88.84%.The prediction results can provide decision support for the highway traffic management department to make more effective safety control measures.

关 键 词：高速公路交通事故 风险预测 改进深度森林算法 深度学习 

分 类 号：X951[环境科学与工程—安全科学]