A Spatio-Temporal Self-Attention Mechanism for Filling and Predicting Missing Values of High-Speed Traffic Flows

Traffic flow prediction is an important research topic in intelligent transport system, and the missing traffic flow data is crucial for traffic flow prediction. However, existing methods seldom consider the problem of missing highway detection data. Therefore, we propose a spatio-temporal self-attention mechanism traffic flow filling and prediction model (FPSTTF) to consider the data loss situation. The model uses the periodicity of traffic flow as the basis and integrates the spatial correlation of road network to fill in the missing values and combines the filling and prediction. The results show that the RMSE and MAE values of the FPSTTF model are reduced by an average of 8.1% and 4.3 % compared to the traditional method, and better filling and prediction results are achieved.