Big data analytics-based traffic flow forecasting using inductive spatial-temporal network

Traffic flow forecasting is crucial for urban traffic management, which alleviates traffic congestion. However, one inherent feature of urban traffic is it's instability, making it difficult to accurately forecast the future traffic flow. In this paper, we propose a model using Inductive Spatial-Temporal Network to predict the traffic flow speed of road networks. Specifically, we first utilize GraphSAGE(Graph SAmple and aggreGatE) to inductively extract the spatial features of road networks. Furthermore, we design a global temporal block to capture the temporal pattern. Then, we adopt the self-attention mechanism for evaluating the importance of nodes. Finally we introduced an autoregressive module to increase the robustness of the model. Experiments on real-world data demonstrate that considering spatial and temporal dependencies of the traffic data can achieves better performance than models without considering such relations.