The invention discloses a short-time traffic congestion prediction method based on a CATST network. The short-time traffic congestion prediction method comprises the following steps: (1) acquiring a real-time road condition map of an open source website map and preprocessing the real-time road condition map; (2) constructing a space-time network, namely a CATST network, based on a convolution attention Transform, and inputting the preprocessed data for training; wherein the CATST network comprises two space-time attention modules, two convolution attention Transform space-time modules, two batch standardization layers and a three-dimensional (3D) convolution layer; (3) evaluating the traffic congestion prediction accuracy of the CATST network by using a mean square error and an average absolute percentage error as evaluation indexes; the method is not only suitable for the space structure of the flow traffic data set, but also has the capability of predicting the state of the point in the future according to the state of the input data and the past state.
本发明公开了一种基于CATST网络的短时交通拥塞预测方法,包括以下步骤:(1)获取开源网站地图的实时路况图并进行预处理;(2)构建基于卷积注意力Transformer时空网络即CATST网络,输入经过预处理后的数据进行训练;其中,CATST网络包括:2个时空注意力模块、2个卷积注意力Transformer时空模块、2个批标准化层和3D卷积层;(3)使用均方误差和平均绝对百分比误差作为评价指标,评价CATST网络对于交通拥塞预测的准确度;本发明不仅适用于流量交通数据集的空间结构,而且还拥有根据输入数据的状态和过去状态来预测未来该点的状态的能力。
Short-term traffic congestion prediction method based on CATST network
一种基于CATST网络的短时交通拥塞预测方法
2024-10-11
Patent
Electronic Resource
Chinese
European Patent Office | 2025
|