Investigation of Traffic Flow Forecasting Utilizing the KMeans-GRU Hybrid Model

Accurate and real-time prediction of traffic flow plays a crucial role in enhancing traffic control and management and tackling the persistent traffic congestion problem on road networks. By addressing this issue, we strive to improve the overall traffic flow efficiency. Given the uncertainty and dynamics of traffic flow, this study proposes a traffic flow prediction method that combines cluster analysis and Gated Recurrent Unit (GRU) neural networks. Through in-depth analysis of historical traffic data, this research delves into the dynamic shifts in traffic flow patterns over time. It uncovers the underlying connections between patterns observed in traffic flow data series. It is found that the selection of training data significantly affects the accuracy of the neural network prediction model. Henceforth, we put forth a novel approach for predicting traffic flow known as KMeans-GRU. This technique combines K-Means clustering and GRU neural networks to achieve accurate predictions. Compared to other prediction models, this method exhibits superior accuracy and strong model generalization capabilities.