Intensive lane-changing (LC) events are one of the great causes that make freeway weaving areas become bottlenecks. This study proposes an approach using vehicle trajectory data to investigate the spatiotemporal distributions of the number of LC events, void occupancies, and throughput variations at the freeway weaving area. Firstly, all LC events are extracted from the cleaned dataset and classified into four types according to the LC vehicles’ origin–destination lanes and LC directions. Secondly, the time and space void occupancies are calculated using the kinematic theory. Thirdly, the throughput variations are identified with the oblique N-curve method. Finally, the spatial and temporal distributions of the LC events, void occupancies, and throughput variations are plotted to analyze their characteristics and relationships. The spatial distributions of different types of LC events indicate that most LC events occur at the surrounding area of the on-ramp entrance. Spatial distributions of time void occupancies show that the time void in the original lanes is quite small while that in the target lanes is much larger. Furthermore, the time void occupancies amplify downstream when considering vehicles traveling on the road. By comparing the temporal distributions of LC events, void occupancies, and throughput variations, there is a lag effect between the large value occurrences of space void occupancy and throughput reduction and that of the LC events, which can conclude a causal relationship between LC events and the occurrences of the space void occupancies and throughput reductions.

Network traffic prediction is essential for intelligent resource management in modern transportation systems, but existing methods struggle to simultaneously capture multi-scale temporal patterns, long-range dependencies, and periodic behaviors while maintaining computational efficiency. This paper presents WMF-Traffic (Wavelet-Mamba-Fourier Traffic prediction framework), a novel traffic forecasting approach that synergistically integrates wavelet decomposition, selective state space modeling, and frequency domain processing. The framework introduces four key components: Multi-scale Wavelet Decomposition for hierarchical temporal pattern extraction, Wavelet Traffic Convolution with scale-adaptive mechanisms, Traffic-aware Mamba for efficient long-range dependency modeling, and Fourier Pattern Adjustment for periodic pattern enhancement. WMF-Traffic employs a comprehensive training objective that balances reconstruction accuracy, temporal consistency, and spectral coherence. Extensive experiments on four real-world traffic datasets (METR-LA, PEMS-BAY, PEMS04, PEMS08) demonstrate consistent improvements over state-of-the-art methods, achieving 1.0-1.3% gains in MAE, 0.6-1.1% in RMSE, and 0.2-1.0% in MAPE across different prediction horizons. Ablation studies reveal that Traffic-aware Mamba provides the largest individual contribution (10.2% MAE reduction), while the complete framework achieves up to 27.1% improvement over baseline approaches. The proposed uncertainty-based fusion mechanism further enhances robustness with 3.2-4.1% additional improvements.

The intelligent transformation of ports has become a critical pathway for advancing sustainable regional development in the context of rapid globalization and digitalization. While existing studies primarily focus on evaluating port intelligence, limited attention has been given to its role in fostering regional economic sustainability. To address this gap, this study examines the Yangtze River Delta (YRD) region as a representative case to explore the mechanisms through which port intelligent upgrading influences regional economic systems. An integrated analytical framework is developed in which the Analytic Hierarchy Process (AHP) is used to construct a composite index of port intelligence, and Principal Component Analysis (PCA) is employed to measure regional economic development. Furthermore, a Panel Vector Autoregression (PVAR) model is applied to capture the dynamic interactions between these two systems. The results indicate that port intelligence has increased significantly across the sampled ports, with the Port of Shanghai and the Port of Ningbo–Zhoushan exhibiting particularly notable growth, rising by 28% and 44%, respectively, over the period of 2011–2021. The PVAR results further reveal a sustained upward trend in regional economic development, suggesting a positive and persistent association between port intelligence and regional economic performance. These findings provide both theoretical and empirical support for understanding the role of port intelligent transformation in promoting regional economic sustainability and offer policy-relevant insights for fostering coordinated development between port systems and regional economies.

Left-turning vehicles from minor streets tend to generate hazards when merging into major streets at un-signalized T-intersections. However, the characteristics of such behavior have been studied relatively little. In this study, an eight-hour video was recorded at un-signalized T-intersections in Nanjing. Two kinds of data were collected from the video, including 1) vehicle trajectories, as well as the corresponding speed profiles and the time records; 2) vehicle yielding characteristics. Considering merging points, diverging points, and turning radius, vehicle trajectories were clustered into six groups. A binary logistic model was developed to evaluate how traffic flow characteristics affected the yielding behavior of left-turning vehicles. A multinomial logistic model was proposed to formulate the choices for six trajectory patterns. The binary logistic model indicated that vehicles would stop when meeting a platoon. The regression model results of the multinomial logistic model showed that high mainline speed encourages the abnormal trajectory patterns. It was found that well-designed object markers and islands in the T-intersection are able to reduce illegal driving behavior and improve safety.

Network traffic prediction is essential for intelligent resource management in self-organizing systems, but existing methods struggle to simultaneously capture multi-scale temporal patterns, long-range dependencies, and periodic behaviors while maintaining computational efficiency. This paper presents WMF-Net, a novel traffic prediction framework that synergistically integrates wavelet decomposition, selective state space modeling, and frequency domain processing. The framework introduces four key components: Multi-scale Wavelet Decomposition for hierarchical temporal pattern extraction, Wavelet Traffic Convolution with scale-adaptive mechanisms, Traffic-aware Mamba for efficient long-range dependency modeling, and Fourier Pattern Adjustment for periodic pattern enhancement. WMF-Net employs a comprehensive training objective that balances reconstruction accuracy, temporal consistency, and spectral coherence. Extensive experiments on four real-world traffic datasets (METR-LA, PEMS-BAY, PEMS04, PEMS08) demonstrate consistent improvements over state-of-the-art methods, achieving 1.0-1.3% gains in MAE, 0.6-1.1% in RMSE, and 0.2-1.0% in MAPE across different prediction horizons. Ablation studies reveal that Traffic-aware Mamba provides the largest individual contribution (10.2% MAE reduction), while the complete framework achieves up to 27.1% improvement over baseline approaches. The proposed uncertainty-based fusion mechanism further enhances robustness with 3.2-4.1% additional improvements.