VMD-LSTM Urban Arterial Short-Term Traffic Forecasting Based on Spatio-Temporal Clustering

Due to the increasingly prominent traffic congestion problem on urban roads, real-time traffic signal timing regulation is provided as a strategy to alleviate this issue. This paper takes the short-time traffic flow data at each intersection of Hongzehu main road as an example. The temporal and spatial clusters of short-time traffic flow are used to divide the data set, respectively, and combined with VMD-LSTM for short-time traffic flow prediction. MAE, RMSE, and MAPE are used as evaluation indexes. In the first step, spatio-temporal clustering is performed to obtain sets with similar or identical characteristics; then, VMD-LSTM is applied to the different sets to obtain the prediction values. The results show that the VMD-LSTM model based on spatio-temporal clustering improves the MAE by 89% on average compared to the LSTM model, and the RMSE improves by 88.47% on average compared to the CNN-LSTM model, which proves the superiority of this model algorithm.