MULTI-VIEW SPATIAL-TEMPORAL CONVOLUTION FOR TRAFFIC FLOW FORECASTING: AN APPROACH TO IMPROVE ACCURACY AND EFFICIENCY

The rapid development of information technology has made multiple time series forecasting, especially in the field of traffic flow analysis, increasingly important for big data evaluation. Traffic flow forecasting, acting as a basis for intelligent transportation systems (ITS), has sig-nificant scientific and practical value for urban traffic planning, improving the quality and efficiency of both logistics transport and public transport strategies. In an intelligent transportation system, traffic flow forecasting remains a critical task. However, due to the complex temporal and spatial dependencies inherent in traffic flow sequences, establishing an accurate forecasting model using ring neural networks, graph networks, and transformer models presents significant challenges. To take advantage of convolutional networks, which are known for their high computational efficiency and robust feature extraction capabilities, a novel traffic flow prediction model is proposed that utilizes multi-view spatio-temporal convolution.