摘要
基于知识图神经推荐算法(KGCN)进行危险品道路运输实时风险评估,旨在通过实时评估危险品道路运输风险,及时发现高风险因素,提前调整运输状态,降低危险品道路运输发生风险的概率和避免造成严重的事故后果。构建危险品道路运输事件知识图谱,基于图神经网络(GNN)将该知识图谱嵌入推荐算法,计算风险因素导致风险事件发生的可能性,针对危险品道路运输事件给出个性化的风险评估结果。该方法简化了实时数据处理操作,克服了实时数据稀疏的困难,以54097条货物道路运输报警数据为实例进行验证,得出AUC值稳定在0.83左右,计算结果可靠。
Knowledge graph neural recommendation algorithm(KGCN)is used to assess hazmat road transportation risks in real time,aiming to reduce the probability of risks and avoid terrible accidents by detecting high-risk factors detected timely and adjusting the transportation status in advance based on real-time risk assessment report on hazmat road transportation.A knowledge graph describing hazmat road transportation accidents is constructed and embedded into a recommendation algorithm based on graph neural networks(GNN),so that the probability of risk factors causing an accident is evaluated and a specialized risk assessment is given for each hazardous material road transport incident.The method simplifies real-time data processing operations and overcomes the difficulties of real-time data sparsity.Taking the cargo road transportation alarm data including 54097 items as an example for calculation,the AUC value is calculated to be stable at around 0.83,indicating that the calculation results are reliable.
作者
王占中
兰若冰
杨萌
张书源
WANG Zhanzhong;LAN Ruobing;YANG Meng;ZHANG Shuyuan(College of Transportation,Jilin University,Changchun 130000,China)
出处
《同济大学学报(自然科学版)》
北大核心
2025年第8期1253-1261,共9页
Journal of Tongji University:Natural Science
基金
吉林省自然科学基金(20230101112JC)。
关键词
道路运输
实时风险评估
知识图谱
图神经网络
推荐算法
road transportation
real-time risk assessment
knowledge graph
graph neural network(GNN)
recommendation algorithm
