The invention discloses a vehicle arrival time prediction method based on representation learning, and belongs to the technical field of intelligent traffic. The objective of the invention is to solve the problem of low vehicle arrival time prediction accuracy in the prior art. According to the method, firstly, an SDNE algorithm is adopted to process an intersection type matrix and a supplementary information matrix including traffic flow information and important information points, and then the processed matrix is input into an MLP full-connection neural network to obtain a first representation vector; processing the node information vector by using an LSTM algorithm to obtain a second representation vector; processing the discrete semantic information vector by adopting an xDeepFM algorithm to obtain a third representation vector; and then splicing the first representation vector, the second representation vector and the third representation vector according to channel, performing one-dimensional ResNet convolution operation, processing through a CBAM attention module and an MLP, and outputting predicted vehicle arrival time by the MLP.
基于表示学习的车辆到达时间预测方法,属于智慧交通技术领域。为了解决现有的车辆到达时间预测准确率较低的问题。本发明首先采用SDNE算法对交叉口类型矩阵和包括交通流量信息、重要信息点的补充信息矩阵进行处理,然后输入到MLP全连接神经网络中得到第一表示向量;采用LSTM算法对节点信息向量行处理得到第二表示向量;同时采用xDeepFM算法对离散语义信息向量进行处理,得到第三表示向量;然后将第一表示向量至第三表示向量按channel拼接,进行1维ResNet做卷积操作,再通过CBAM注意力模块和MLP处理,MLP输出预测的车辆到达时间。
Vehicle arrival time prediction method based on representation learning
基于表示学习的车辆到达时间预测方法
2024-04-19
Patent
Electronic Resource
Chinese
European Patent Office | 2018
|ARRIVAL TIME PREDICTION METHOD, ARRIVAL TIME PREDICTION DEVICE, AND ARRIVAL TIME PREDICTION PROGRAM
European Patent Office | 2022
|ARRIVAL TIME PREDICTION DEVICE AND ARRIVAL TIME PREDICTION DEVICE SYSTEM
European Patent Office | 2017
|