Deep learning based traffic flow prediction model on highway research

With the acceleration of urbanization and the continuous improvement of per capita vehicle ownership rate, traffic congestion and traffic accidents have become a global problem, especially highway congestion has become a national problem. The expressway is a bridge to the modernization of a country, and it is the only way to develop the modern transportation industry. The traffic congestion phenomenon on the expressway not only brings great inconvenience to people's travel, but also restricts the improvement of the service quality of the expressway, and affects the regional economic and social development. Predicting highway traffic flow one or more days in advance can not only assist highway management personnel to arrange deployment in advance, reasonably induce vehicle diversion and evacuation, but also provide reference for the public to select travel routes in advance, and is also an effective way to alleviate highway congestion. Aiming at the problems of low fine reading, poor real-time performance, insufficient adaptability and robustness of the current traffic flow prediction model in the field of intelligent traffic prediction, this study proposed an improved graph convolutional neural network for traffic flow prediction in intelligent traffic prediction. The model is composed of GCN and LSTM. Self- attention module is embedded in GCN backbone network to improve the ability of network to extract spatial features. The ECA attention module is embedded in the backbone network of LSTM to improve the ability of the network to extract time features.