Short-Term Traffic Speed Forecasting Using a Deep Learning Method Based on Multitemporal Traffic Flow Volume

Accurate traffic speed forecasting not only can help management departments make better judgments and improve the efficacy of road monitoring but also drivers plan their driving routes arrive safely smoothly at destination. This paper focuses on lack data proposes a method for based multitemporal flow volume previous later moment states. First, according to data, different patterns states were extracted. Second, performance five models, namely, long short-term memory (LSTM), backpropagation (BP), classification regression trees, k-nearest neighbor, support vector regression, compared. Finally, model with best prediction results was used conduct sensitivity analysis experiments patterns. Through real-data case study, we found that LSTM has highest accuracy compared other models in both time space. pattern “previous = 3 3” forecast more accurately, its ability is robust across range scenarios.