Urban expressway traffic state recognition method based on Dynamic Step Firefly and Fuzzy C-Means clustering

Accurate real-time traffic state recognition is crucial for intelligent transportation systems, as it enables proactive traffic management and congestion alleviation. However, achieving high recognition accuracy remains challenging due to the complexity and volatility of urban expressway traffic. This paper proposes a novel traffic state recognition method that integrates a Dynamic Step Firefly Algorithm (DSFA) with Fuzzy C-Means (FCM) clustering to address this challenge. The traffic states are reclassified based on existing standards, and three fundamental parameters—cross-sectional traffic flow, cross-sectional average speed and cross-sectional average time occupancy rate—are selected as state variables. The DSFA is innovatively applied to optimize the initial clustering centers for the FCM algorithm, thereby mitigating its sensitivity to initialization. The model is rigorously evaluated using the silhouette coefficient as a clustering quality metric. Experimental results on loop detector data from an urban expressway, collected over five consecutive days (August 27-31, 2018) show that the proposed DSFA-FCM method achieves a high global silhouette coefficient of approximately 0.9, which outperforms the Firefly Algorithm-optimized FCM (FA-FCM) by 0.25 and the baseline FCM by 0.452. These quantitative improvements confirm that the dynamic step-size mechanism significantly enhances global search efficiency and clustering robustness. The key contribution of this research is the demonstration that dynamically optimizing the FCM's initial centers via DSFA significantly enhances recognition accuracy, offering a robust solution for practical traffic management applications.