The invention provides a short-time traffic flow prediction method based on an integrated graph convolutional neural network, and belongs to the field of intelligent traffic, the integrated graph convolutional neural network and a convolutional neural network are combined on processing graph data and a traffic graph, firstly, the traffic graph comprises a complex road network topology, and the road network topology is a complex road network topology; the GCN is used for extracting features through information of a current node and a neighbor node, the feature extraction effect is enhanced, secondly, for traffic map data, the CNN can extract local features of an interested area, more detailed traffic flow prediction data are provided, multiple linear layers are used for fusing features of different levels, the complex traffic data relation can be better processed, and the traffic flow prediction efficiency is improved. In order to adapt to output features, a special loss function is designed to more accurately evaluate a model output result, so that the accuracy of flow prediction is improved, and due to the innovation points, the prediction accuracy and robustness can be improved, and the complexity of traffic data can be dealt with.
本发明提供了一种基于集成图卷积神经网络的短时交通流预测方法,属于智能交通领域,在处理图数据和交通图上结合集成图卷积神经网络和卷积神经网络,首先,由于交通图包括复杂的道路网络拓扑,故使用GCN通过当前节点与邻居节点的信息提取特征,增强了特征提取的效果,其次,对于交通图数据,CNN则可提取感兴趣区域的局部特征,提供更详细的交通流预测数据,采用多层线性层融合不同层级特征,可更好地处理复杂的交通数据关系,提高深层次特征提取,实现更精确的流量预测,为适应输出特征,设计了专门的损失函数,以更准确地评估模型输出结果,从而提高流量预测的精确性,这些创新点可提高预测准确性、鲁棒性,并应对交通数据的复杂性。
Short-term traffic flow prediction method based on integrated graph convolutional neural network
一种基于集成图卷积神经网络的短时交通流预测方法
2024-02-23
Patent
Electronic Resource
Chinese
Short-term traffic flow prediction method based on graph convolution recurrent neural network
European Patent Office | 2024
|