为提升在非线性交通场景中的适应性,有效捕捉交通流的时空关联特征,本文提出融合RBF神经网络与斑马优化算法(ZOA)的短时交通量预测模型。首先采集四向交通流量数据,经去噪归一化后,利用斑马优化算法(ZOA)动态划分拥堵敏感区域并构建邻...为提升在非线性交通场景中的适应性,有效捕捉交通流的时空关联特征,本文提出融合RBF神经网络与斑马优化算法(ZOA)的短时交通量预测模型。首先采集四向交通流量数据,经去噪归一化后,利用斑马优化算法(ZOA)动态划分拥堵敏感区域并构建邻接矩阵;将区域流量输入RBF神经网络预测未来流量,同时将预测误差反馈至ZOA优化信号控制策略;实测显示区域通行效率提升19.3%,预测误差(RMSE 9.887%、MAE 7.976%、RMSPE 11.017%和MAPE 11.735%)显著低于传统模型。该模型通过实时优化网络参数,有效捕捉交通流的时空关联特征,为动态交通管控提供了高精度的预测支持,对缓解城市交通拥堵具有重要实践价值。In order to improve the adaptability in nonlinear traffic scenarios and effectively capture the spatiotemporal correlation characteristics of traffic flow, this paper proposes a short-term traffic volume prediction model based on RBF neural network and Zebra Optimization Algorithm (ZOA). Firstly, the four-way traffic flow data was collected, and after denoising and normalization, the Zebra Optimization Algorithm (ZOA) was used to dynamically divide the congestion sensitive areas and construct an adjacency matrix. The regional traffic is input into the RBF neural network to predict the future traffic, and the prediction error is fed back to the ZOA optimization signal control strategy. The measured results show that the regional traffic efficiency is increased by 19.3%, and the prediction error (RMSE 9.887%, MAE 7.976%, RMSPE 11.017% and MAPE 11.735%) is significantly lower than that of the traditional model. By optimizing network parameters in real time, the model effectively captures the spatiotemporal correlation characteristics of traffic flow, provides high-precision prediction support for dynamic traffic control, and has important practical value for alleviating urban traffic congestion.展开更多
文摘为提升在非线性交通场景中的适应性,有效捕捉交通流的时空关联特征,本文提出融合RBF神经网络与斑马优化算法(ZOA)的短时交通量预测模型。首先采集四向交通流量数据,经去噪归一化后,利用斑马优化算法(ZOA)动态划分拥堵敏感区域并构建邻接矩阵;将区域流量输入RBF神经网络预测未来流量,同时将预测误差反馈至ZOA优化信号控制策略;实测显示区域通行效率提升19.3%,预测误差(RMSE 9.887%、MAE 7.976%、RMSPE 11.017%和MAPE 11.735%)显著低于传统模型。该模型通过实时优化网络参数,有效捕捉交通流的时空关联特征,为动态交通管控提供了高精度的预测支持,对缓解城市交通拥堵具有重要实践价值。In order to improve the adaptability in nonlinear traffic scenarios and effectively capture the spatiotemporal correlation characteristics of traffic flow, this paper proposes a short-term traffic volume prediction model based on RBF neural network and Zebra Optimization Algorithm (ZOA). Firstly, the four-way traffic flow data was collected, and after denoising and normalization, the Zebra Optimization Algorithm (ZOA) was used to dynamically divide the congestion sensitive areas and construct an adjacency matrix. The regional traffic is input into the RBF neural network to predict the future traffic, and the prediction error is fed back to the ZOA optimization signal control strategy. The measured results show that the regional traffic efficiency is increased by 19.3%, and the prediction error (RMSE 9.887%, MAE 7.976%, RMSPE 11.017% and MAPE 11.735%) is significantly lower than that of the traditional model. By optimizing network parameters in real time, the model effectively captures the spatiotemporal correlation characteristics of traffic flow, provides high-precision prediction support for dynamic traffic control, and has important practical value for alleviating urban traffic congestion.
