Spatiotemporal Parallel Convolutional Network Considering Multidimensional Factors for Short-Term Traffic Flow Prediction

This research proposes a spatiotemporal parallel convolutional network (SPCN) model that accounts for both the interactions between lanes and the impact of weather conditions in predicting short-term traffic volume. Initially, multidimensional features, including traffic volume and weather data, are used as the inputs to the SPCN model. To capture the temporal and spatial characteristics of these features, a parallel convolutional block (PCB) based on convolutional neural networks is designed. The SPCN model is then built upon the PCB to predict traffic volumes. Experimental results demonstrate that the SPCN model outperforms other models in terms of prediction accuracy. This confirms that considering both lane interactions and weather influences leads to more accurate and effective traffic volume predictions.