Research on the Prediction Model of Urban Road Congestion Index Based on Gaussian Nearest Neighbors

Against the backdrop of an accelerating urbanization process, traffic congestion has become a significant factor restricting urban economic development and affecting the quality of life of residents. To effectively alleviate traffic pressure during peak hours and reduce energy consumption and environmental pollution, this paper proposes a short-term prediction model for urban road congestion index based on the Gaussian Nearest Neighbor (GNN) algorithm. This model deeply explores the feature information of historical traffic state sequences, measures the similarity between the target sequence and historical sequences using Euclidean distance, and utilizes the Gaussian distribution function to weight and allocate similar sequences, thereby achieving dynamic weighted prediction of the optimal neighbor sequence. The data undergoes a systematic preprocessing process including standardization, sliding window modeling, and feature selection to construct a high-quality dataset suitable for model training. In the 2-minute short-term prediction task, the model demonstrates excellent prediction performance, with a Mean Absolute Error (MAE) of .23, a Mean Root Square Error (MRSE) of .33, and a Mean Absolute Percentage Error (MAPE) of only 4.70%. The predicted values show a highly consistent trend with the actual values. The experimental results indicate that the GNN model has high prediction accuracy and good real-time response capability in short-term traffic congestion index prediction tasks, and possesses strong scalability, providing strong technical support for the optimization and decision-making of urban intelligent transportation management systems.