In the current day, one of the main causes of the significant negative effects that urban areas have on the environment and the economy globally is traffic congestion. Predicting traffic flow is a critical component of the intelligent transportation system. Anticipating future traffic is one of the most effective methods. It is possible to reduce traffic and make travel safer and more affordable by studying traffic predictions. The deep neural network era acquired prominence because of its extraordinary prediction capacity, which is a result of its intricate and profound structure. Even though deep neural network models are widely used in traffic prediction, there isn’t much literature on these techniques. Our paper provides an overview of some of the most recent research on traffic flow prediction using deep learning and spatial-temporal deep learning. Numerous deep learning configurations consist of graph convolutional neural networks (GCNN), multi-graph convolutional networks (MGCN), dynamic graph convolutional networks (DGCN), and spatial temporal graph convolutional networks (STGCN). Many layers are used by these deep learning models to gradually extract higher-level properties from the initial data. An updated survey of deep neural networks for traffic prediction is presented in this paper. Given the intricacy of transport systems, the most recent deep learning models created to address this particular topic are examined, along with information on how different aspects affect the models and which models perform best in specific settings. We have also provided future research problems and emergent challenges at the end of the paper.
Sridhar et al. (Tue,) studied this question.