Spatio-Temporal CONV-LSTM Traffic Flow Prediction Algorithm

Traditional traffic flow prediction methods often focus on temporal or spatial information in isolation, overlooking the correlation between spatio-temporal features. This paper proposes a regional traffic density prediction algorithm based on spatio-temporal information, named the Convolutional Long Short-Term Memory (CONV-LSTM) model. The algorithm begins by constructing a regional traffic flow matrix using mathematical modeling, which decomposes the complex prediction problem into more manageable sub-problems. Subsequently, Convolutional Neural Networks (CNNs) and Long Short- Term Memory Networks (LSTMs) extract spatial and temporal features from the dataset, thereby integrating spatio- temporal correlations for a comprehensive analysis. Experimental validation indicates that the algorithm effectively reduces the Root Mean Square Error (RMSE) to approximately 12% on both workdays and non-workdays and achieves a lower Mean Absolute Error (MAE) of about 8%. These results underscore the algorithm's effectiveness in accurately forecasting regional traffic flow, with significant practical implications for traffic engineers, urban planners, and transportation and traffic management researchers.