S-GCN-GRU-NN: A novel hybrid model by combining a Spatiotemporal Graph Convolutional Network and a Gated Recurrent Units Neural Network for short-term traffic speed forecasting

Forecasting the short-term speed of moving vehicles plays an important role not only in reducing travel time, but also in saving energy and reducing air pollution. However, it still remains a challenging task when the high accuracy is required. In this paper, we propose a novel hybrid model named S-GCN-GRU-NN, in which a novel spatiotemporal graph convolutional network (S-GCN) model is proposed for acquiring the complex spatiotemporal dependence, and a gated recurrent units neural network (GRU-NN) model is used for short-term traffic speed forecasting. The extensive experimental results show that, the proposed hybrid model has higher stability and accuracy than other models, including S-GCN model, GRU-NN model, autoregressive integrated moving average (ARIMA) model, support vector regression (SVR) model, k-nearest neighbor (KNN) model, multi-layer perceptron (MLP) model and long short-term memory neural network (LSTM-NN) model. In addition, we find that the time lag is a key effect factor for the model performances.