The invention discloses a traffic flow prediction method based on multi-model network coupling. The method comprises the following steps: processing and inputting road traffic flow original data; inputting the data into a linear layer for preprocessing to obtain linear transformation; inputting into Encoder, and extracting spatial dependency and time dependency; inputting an output result of the encoder into AdjostBlock to extract local features and enhance context information; summing the output result of the AdjostBlock and the original data to obtain a residual error, and taking the residual error as the input data of Decoder to carry out subsequent learning; an output result of the decoder is used as Query data of cross attention, and an output result of the encoder is used as Key data and Value data to be sent into the cross attention to process complex sequence data and then is input into a full connection layer to obtain a final prediction result. According to the invention, the accuracy of traffic flow prediction data prediction can be improved.
本发明公开了一种基于多模型网络耦合的交通流量预测方法,包括:道路交通流量原始数据的处理和输入;将数据输入线性层进行预处理得到线性变换;输入进Encoder,提取空间依赖性和时间依赖性;将编码器的输出结果输入AdjustBlock提取局部特征和增强上下文信息;将AdjustBlock的输出结果和原始数据进行加和残差,作为Decoder的输入数据进行后续学习;解码器的输出结果作为交叉注意力的Query数据,编码器的输出结果作为Key数据和Value数据送入交叉注意力中处理复杂的序列数据后输入到全连接层得到最后的预测结果。本发明可以提升交通流量预测数据预测的准确性。
Traffic flow prediction method based on multi-model network coupling
一种基于多模型网络耦合的交通流量预测方法
2024-09-20
Patent
Electronic Resource
Chinese
European Patent Office | 2025
|Traffic speed prediction method based on multi-moment traffic flow
European Patent Office | 2025
|Complex traffic network flow prediction method based on deep learning model
European Patent Office | 2024
|Traffic flow prediction method based on multi-view dynamic graph convolutional network
European Patent Office | 2022
|