Track trials, findings, and clinical evidence as it lands.
Search papers, labs, and topics across Lattice.
Recent developments of Artificial Intelligence of Things (AIoT) are transforming several sectors like Industry 4.0, smart cities, intelligent healthcare, and smart transportation. The AIoT can be employed in urban mobility to enable real-time monitoring and intelligent decision-making in the transportation sector. Traffic congestion classification is a significant aspect of Intelligent Transportation Systems (ITS), as the rising urban population has contributed to severe congestion complexity that impacts air quality, road safety, and travel time. Accurate classification of congestion patterns improves traffic management and mitigates environmental impacts. Although precisely predicting congestion spread is still problematic, owing to the complex, non-linear, and dynamic nature of traffic patterns. As deep learning advances, efficient models are required to address the challenges in smart urban mobility systems. Therefore, this study presents a Deep Learning-Based Urban Mobility Intelligent System for Traffic Congestion Framework (DLUMIS-TCF). The proposed DLUMIS-TCF model analyzes urban mobility data with trip-level records for precise traffic congestion level classification. The preprocessing stage includes missing value handling, outlier detection, feature scaling, temporal attribute extraction, and spatial coordinate utilization to prepare the data for effective learning. For feature selection, the model utilizes Elastic Net to select the relevant spatio-temporal features from the dataset. In addition, the traffic congestion level classification is performed using a heterogeneous spatiotemporal graph attention network. Furthermore, the optimization of the model is carried out using the AdaBelief optimizer to improve convergence and performance. To improve the interpretability of the model, SHAP is used to explain the contribution of input features toward the classification results. To assess the performance of the proposed DLUMIS-TCF model, experiments are carried out on an Urban Mobility Intelligence Dataset obtained from the Kaggle repository. The experimental results show the effectiveness of the proposed model in predicting traffic congestion compared to other methods. Thus, the proposed DLUMIS-TCF approach can be useful for urban traffic management, congestion control, and intelligent transportation systems.