Traffic Flow Predication by Using Graph Convelution Network

These days, a wide range of applications and electrical gadgets employ machine learning techniques to solve issues and forecast future conditions. Since there are so many cars on the road and traffic congestion is readily caused by them, one of the problems that most large cities face is traffic jams. Although traffic congestion are bad for the environment, their effects may be reduced with careful planning. Because it may prevent traffic congestion by using its predictive knowledge, one of the most intriguing topics for intelligent transportation systems is traffic prediction. It is quite difficult for academics to develop or use a model that will function properly in various conditions when it comes to traffic prediction. One of the most important aspects of traffic prediction is capturing both geographical and temporal connections. Taking use of model combinations is one approach to capturing both dependencies. the model combine three layer the graph convolution network and the long short term memory . Graph convolutional network is used by the model to learn the spatial dependency based on road network architecture, and long short term memory are used to learn the short-time trend in time series. To further increase prediction accuracy, the attention mechanism was included to modify the weight assigned to various time points and compile global temporal information.