A Long-term Traffic Flow Prediction Model Based on Multiple Features

Aiming to solve the problem of low prediction performance in long-term traffic flow prediction, we suggest a new forecasting method called Multiple Feature Temporal Graph Convolutional Networks (MF-TGCN). In long-term traffic flow prediction, some characteristics beyond road conditions can also become the main factors affecting traffic flow changes. In this article, we take into account factors such as weather, temperature, and holidays in road network traffic, extract spatial features of traffic information through Graph Convolutional Networks (GCN), and extract temporal features of traffic information through Gate Recurrent Unit(GRU). At the same time, to improve the prediction accuracy of the model, we used a threshold Gaussian kernel to calculate the weight of roads between adjacent stations. On the basis of real traffic flow information, the results show that the MF-TGCN model performence better than other current prediction model.