The invention relates to the technical field of computers, and discloses a road traffic flow prediction model construction method and device, a road traffic flow prediction method and device, equipment and a medium, and the method comprises the steps: obtaining a first traffic flow sample data set corresponding to a preset road type; inputting the position data and the time data into a pre-configured physical constraint neural network model, and performing iterative training on the model to obtain a total loss function value; and when the total loss function value meets a preset iteration stopping condition, stopping iteration training, and obtaining an optimal physical constraint network model. The model combines the neural network model, the preset type traffic flow equation corresponding to the preset type principle and the physical constraint, can better adapt to different traffic conditions and road conditions, has certain generalization ability, and improves the universality and reliability of the road traffic flow prediction model.
本发明涉及计算机技术领域,公开了道路交通流预测模型构建、预测方法、装置、设备及介质,该方法包括:获取与预设道路类型对应的第一交通流样本数据集;将位置数据和时间数据,输入至预配置的物理约束神经网络模型中,对该模型进行迭代训练,获取总损失函数值;当总损失函数值符合预设停止迭代条件时,停止迭代训练,获取最优的物理约束网络模型。该模型中结合了神经网络模型、与预设类型道理对应的预设类型交通流方程和物理约束,可以更好地适应不同的交通情况和道路条件,具有一定的泛化能力,提高道路交通流预测模型的通用性和可靠性。
Road traffic flow prediction model construction method and device, road traffic flow prediction method and device, equipment and medium
道路交通流预测模型构建、预测方法、装置、设备及介质
2024-10-15
Patent
Electronic Resource
Chinese
Road traffic flow prediction method and system, medium and equipment
European Patent Office | 2020
|Urban road traffic flow prediction method, device and equipment and storage medium
European Patent Office | 2024
|