Optimizing Urban Mobility: Advancing Traffic Prediction and Transportation Analytics in Complex Networks

Traffic congestion remains a critical challenge in urban areas, especially in high-density zones like Bengaluru, India. This study proposes an integrated traffic management system incorporating real-time vehicle classification and traffic flow prediction to address this problem. Using YOLOv8 for vehicle classification, seven distinct vehicle categories were identified with accuracy above 70 percent. The system uses XG-Boost for traffic flow prediction, outperforming traditional models by capturing non-linear traffic patterns. Video data from 23 surveillance cameras were used to create a data set for vehicle detection and flow analysis. Results showed that YOLOv8 performed better in detecting small vehicles in complex urban scenarios, while XGBoost achieved a mean squared error of 0.62, which is significantly better than other models. An integrated approach that facilitates real-time monitoring and prediction for enhancing urban traffic management through better decision-making and mitigation of congestion is developed. The solution has been scalable and robust, allowing for its application in smart city traffic management systems.