A deep residual long short-term memory network model for traffic flow prediction

Abstract

The scale of traffic is rising as a result of rapid economic development, and there are more and more vehicles on the road, which can easily cause road traffic congestion and even accidents. Accurate traffic flow prediction has been a research focus of intelligent transportation systems. Taking emergency steps for prevention and management ahead of time is critical to traffic safety and transportation efficiency. Existing models and methods lack the ability to model both temporal and spatial correlations, and suffer from poor extraction of temporal and spatial features, which leads to reduced prediction performance. In this paper, a fusion model SECLSTM is proposed to implement traffic flow prediction. The model includes long short-term memory (LSTM) network, convolutional residual network and Squeeze-and-Excitation (SE) module. A convolutional residual network based on the attention mechanism squeeze excitation module is constructed and fused with the LSTM network. Experiments are carried out using real data sets, and the findings reveal that the proposed model outperforms existing methods in terms of accuracy and efficiency.