Adaptive Cascade Clustering for High-Fidelity Urban Traffic Pattern Recognition

Inefficiencies in urban traffic management, particularly from static traffic light regulation, pose significant challenges to optimizing traffic flows and mitigating environmental impact. Existing analytical methods often lack the adaptability to autonomously detect the complex, dynamic structure of traffic patterns. In this work, we introduce an adaptive cascade clustering approach that synergizes Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) and k-means algorithms. By employing a weighted voting mechanism, our approach integrates the benefits of density-based structural analysis with centroidal cluster refinement, advancing upon existing hybrid models. Evaluated on a high-fidelity simulation model of the Khmelnytskyi transport network in Khmelnytskyi, Ukraine, the proposed approach demonstrated a superior ability to identify true traffic modes. It achieved a V-measure of 0.79–0.82 and improved cluster compactness by 4–13 % compared to standalone algorithms. Furthermore, the model attained a scenario identification accuracy of 92.8–95.0 % with a temporal coherence of 0.94. These findings confirm that by leveraging adaptive cascade principles, our approach significantly enhances the quality of traffic mode identification, representing a key advancement for developing more intelligent and responsive urban transport management systems.