Smart Traffic Flow Prediction Using Deep Learning: A Systematic Review of Recent Advances, Challenges, and Future Research Directions
TL;DR
This systematic review critically examines recent advancements in deep learning-based traffic flow prediction models, emphasizing studies published between 2018 and 2026 and indicates that although graph-based and transformer-based architectures currently achieve state-of-the-art predictive performance, integrating explainability, real-time adaptation, and privacy-preserving learning remains a significant research priority.
Abstract
Rapid urbanization, population growth, and the increasing number of vehicles have significantly intensified traffic congestion across metropolitan regions worldwide. Conventional traffic management systems are often inadequate for addressing the dynamic and nonlinear nature of urban transportation networks. Consequently, artificial intelligence (AI), particularly deep learning (DL), has emerged as a transformative approach for intelligent traffic flow prediction. Accurate traffic forecasting enables proactive traffic management, optimized route planning, reduced travel time, lower fuel consumption, and improved road safety, thereby contributing to the development of smart and sustainable cities. This systematic review critically examines recent advancements in deep learning-based traffic flow prediction models, emphasizing studies published between 2018 and 2026. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology, this review synthesizes findings from high-quality journal articles, conference proceedings, and technical reports indexed in Scopus and Web of Science. The paper evaluates major deep learning architectures, including Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Networks (CNN), Graph Neural Networks (GNN), Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), Transformer-based architectures, and hybrid deep learning models. The review further analyzes their predictive performance, computational efficiency, scalability, interpretability, and real-world applicability in intelligent transportation systems. Existing challenges such as data heterogeneity, missing sensor data, privacy concerns, model explainability, computational cost, and deployment limitations are critically discussed. A conceptual research framework highlighting emerging technologies—including edge computing, Internet of Things (IoT), digital twins, federated learning, explainable AI, and large foundation models—is proposed to guide future research. The review contributes theoretically by synthesizing fragmented literature, technologically by identifying advanced predictive architectures, managerially by providing recommendations for transportation authorities, and sustainably by demonstrating how AI-driven traffic prediction supports greener and more efficient urban mobility. The findings indicate that although graph-based and transformer-based architectures currently achieve state-of-the-art predictive performance, integrating explainability, real-time adaptation, and privacy-preserving learning remains a significant research priority. Keywords: Smart Traffic Prediction; Deep Learning; Intelligent Transportation Systems; Graph Neural Networks; Traffic Forecasting; Explainable Artificial Intelligence; Smart Cities