Open research questions in Traffic Prediction and Management Techniques

What the literature leaves open

• The need to capture the complex spatial-temporal dependencies of inbound and outbound flow from highway toll stations during holidays. The challenge of achieving accurate, reliable traffic flow predictions during holidays when surging traffic poses significant challenges.

  Multi-Task Learning-Based Traffic Flow Prediction Through Highway Toll Stations During Holidays · 2025 · DOI

• Existing studies have not effectively captured the complex spatial-temporal dependencies of inbound and outbound flow from highway toll stations during holidays. There is a need for a joint prediction model that can accurately predict both inbound and outbound flow. The lack of a multi-task learning framework that can share spatial–temporal features between inbound flow and outbound flow.

  Multi-Task Learning-Based Traffic Flow Prediction Through Highway Toll Stations During Holidays · 2025 · DOI

• The time span of the bike-sharing data does not exactly match that of the taxi and bus trajectory data. The analysis is limited to the same day from the same quarter to minimize the interference of seasonal factors.

  A Spatiotemporal Multi-Model Ensemble Framework for Urban Multimodal Traffic Flow Prediction · 2025 · DOI

• Existing methods have limitations in capturing diverse trajectory characteristics. Prior work does not effectively integrate and analyze all modes of transportation. There is a need for a novel framework that can capture dynamic interactions between multimodal travel trajectories.

  A Spatiotemporal Multi-Model Ensemble Framework for Urban Multimodal Traffic Flow Prediction · 2025 · DOI

• Further research could investigate the use of other machine learning models for time series forecasting. The application of Bayesian networks for time series forecasting could be explored. The development of new methods to deal with high-dimensional big data could be a future research direction.

  Identify Optimal Pedestrian Flow Forecasting Methods in Great Britain Retail Areas: A Comparative Study of Time Series Forecasting on a Footfall Dataset · 2025 · DOI

• The issue of trajectory points from different branches being in close proximity, causing interference in the tracking process. The need to determine the optimal radius of the sliding window and step size to ensure accurate tracking. The challenge of distinguishing automobile trajectories from other types of trajectories, such as pedestrian walking and cycling.

  Automatic Extraction of Road Interchange Networks from Crowdsourced Trajectory Data: A Forward and Reverse Tracking Approach · 2025 · DOI

• The lack of comprehensive CCTV coverage in urban back streets creates gaps in observational data, limiting the ability to capture congestion patterns. Conventional models are not designed to handle sparse and partially observed data, resulting in reduced accuracy and efficiency. There is a need for a lightweight and efficient model that can infer congestion in unmonitored segments.

  PGTFT: A Lightweight Graph-Attention Temporal Fusion Transformer for Predicting Pedestrian Congestion in Shadow Areas · 2025 · DOI

• The need to improve parameter effectiveness while maintaining forecasting accuracy is a challenge. The paper also mentions the challenge of extracting bidirectional dependence in traffic flow series.

  STAE-BiSSSM: A Traffic Flow Forecasting Model with High Parameter Effectiveness · 2025 · DOI

• The lack of standardized interfaces and protocols for integrating plug-and-play artificial intelligence models. The limitations of traditional physical models and static objects in existing geographic data analysis software. The need for new technology to handle the computation of large datasets and provide AI-based calculation methods.

  STPam: Software for Intelligently Analyzing and Mining Spatiotemporal Processes Based on Multi-Source Big Data · 2025 · DOI

• Graph Attention Network (GAT) methods are hindered by prohibitive computational costs. The need for a lightweight and efficient model for traffic violation prediction. The lack of a model that can preserve modeling capability while achieving significant efficiency gains.

  LTVPGA: Distilled Graph Attention for Lightweight Traffic Violation Prediction · 2025 · DOI

• Static embedding fusion cannot dynamically capture semantic importance variations during denoising. The lack of a semantically conditioned framework for short-term OD flow forecasting.

  Cross-Attention Diffusion Model for Semantic-Aware Short-Term Urban OD Flow Prediction · 2025 · DOI

• Previous research focused on single primary disturbances, while this study examines multiple primary disturbances. The lack of a comprehensive framework for assessing the impact of AI optimization on operational resource use and cost-benefit analysis. The need for a more efficient and effective approach to improve headway regularity and reduce disturbances in urban public transportation systems.

  AI-Driven Approach for Enhancing Sustainability in Urban Public Transportation · 2024 · DOI

• Traditional methods for short-term traffic flow prediction have limitations in capturing complex patterns in traffic flow data. A single model such as GRU may not suffice to meet the requirements of intricate time series data.

  Convolutional Neural Network-Based Bidirectional Gated Recurrent Unit–Additive Attention Mechanism Hybrid Deep Neural Networks for Short-Term Traffic Flow Prediction · 2024 · DOI

• The paper does not explicitly identify specific technical or domain challenges, but mentions the complexity of issues in intelligent transportation. The integration of ML and DL in intelligent transportation poses challenges. The paper aims to address these challenges by providing a comprehensive overview of current research trends and future directions.

  The Application of Machine Learning and Deep Learning in Intelligent Transportation: A Scientometric Analysis and Qualitative Review of Research Trends · 2024 · DOI

• Existing methods rely on a fixed-graph structure, which can restrict the representation of spatial information. The fixed-graph structure cannot capture dynamic spatial correlations due to varying conditions such as time and road changes.

  Mixed-Graph Neural Network for Traffic Flow Prediction by Capturing Dynamic Spatiotemporal Correlations · 2025 · DOI

• Traditional time series forecasting technology has limited ability to deal with high-dimensional big data and effectively express complex functions. The literature about classic methods is vast. No sensor is installed around Notting Hill tube station.

  Identify Optimal Pedestrian Flow Forecasting Methods in Great Britain Retail Areas: A Comparative Study of Time Series Forecasting on a Footfall Dataset · 2025 · DOI

• Previously proposed traffic flow prediction methods have high computational complexity, elevated memory usage, and model overfitting. Existing methods for constructing spatial dependencies perform fine-grained calculations across the entire traffic graph, leading to substantial complexity and memory requirements for large-scale traffic networks.

  Grid Partition-Based Dynamic Spatial–Temporal Graph Convolutional Network for Large-Scale Traffic Flow Forecasting · 2025 · DOI

• The limited representation of node transmission capabilities and distance-sensitive interactions in road networks. The inability to capture temporal dynamics in spatial dependencies within traffic flow.

  A Traffic Flow Forecasting Method Based on Transfer-Aware Spatio-Temporal Graph Attention Network · 2025 · DOI

• Investigate the impact of batch size on other deep learning models. Explore the use of other optimization techniques to find the optimal batch size. Consider other factors that may affect the performance of deep learning models.

  Determination of Optimal Batch Size of Deep Learning Models with Time Series Data · 2024 · DOI

• Traditional methods are labor-intensive, time-consuming, and limited in coverage - Field surveys are difficult to scale - Manual observations lack consistency and are prone to subjectivity - Questionnaires may suffer from small sample sizes and non-representative…

  Automated Identification and Spatial Pattern Analysis of Urban Slow-Moving Traffic Bottlenecks Using Street View Imagery and Deep Learning · 2025 · DOI

• Integration of temporal imagery updates, transfer learning, and collaborative training - Application of deep learning in mobility research - Investigation of spatial effects analysis of slow-moving traffic…

  Automated Identification and Spatial Pattern Analysis of Urban Slow-Moving Traffic Bottlenecks Using Street View Imagery and Deep Learning · 2025 · DOI

• Future research can focus on improving the model's performance on small-scale traffic graphs, - Future research can explore the use of other methods to capture global traffic conditions, - Future research can investigate the use of transfer learning to reduce the amount of training data…

  Grid Partition-Based Dynamic Spatial–Temporal Graph Convolutional Network for Large-Scale Traffic Flow Forecasting · 2025 · DOI

• Further research could explore the application of PGTFT in other domains, - Investigating the use of PGTFT in emergency response and transportation planning could be a promising…

  PGTFT: A Lightweight Graph-Attention Temporal Fusion Transformer for Predicting Pedestrian Congestion in Shadow Areas · 2025 · DOI

• Further research can focus on improving the scalability of CADM for larger transportation networks, - Investigating the application of CADM in other domains such as traffic signal…

  Cross-Attention Diffusion Model for Semantic-Aware Short-Term Urban OD Flow Prediction · 2025 · DOI

• Incorporate spatiotemporal correlation analysis methods to dynamically analyze the evolution patterns of crash hazard hotspots, - provide more accurate support for the development of traffic management…

  Optimizing Kernel Density Estimation Bandwidth for Road Traffic Accident Hazard Identification: A Case Study of the City of London · 2024 · DOI

Most-cited papers in Traffic Prediction and Management Techniques

• T-GCN: A Temporal Graph Convolutional Network for Traffic Prediction · IEEE Transactions on Intelligent Transportation Systems · 2020 · 3,160 citations
• Extraordinary Rendition and the Constitution: The Case of Maher Arar · ˜The œReview of litigation · 2008 · 557 citations
• A Hybrid Deep Learning Model With Attention-Based Conv-LSTM Networks for Short-Term Traffic Flow Prediction · IEEE Transactions on Intelligent Transportation Systems · 2021 · 499 citations
• Predicting citywide crowd flows using deep spatio-temporal residual networks · Artificial Intelligence · 2018 · 478 citations
• Stacked bidirectional and unidirectional LSTM recurrent neural network for forecasting network-wide traffic state with missing values · Transportation Research Part C Emerging Technologies · 2020 · 445 citations
• Exploiting dynamic spatio-temporal graph convolutional neural networks for citywide traffic flows prediction · Neural Networks · 2021 · 427 citations
• Short-Term Traffic Flow Prediction for Urban Road Sections Based on Time Series Analysis and LSTM_BILSTM Method · IEEE Transactions on Intelligent Transportation Systems · 2021 · 396 citations
• Traffic transformer: Capturing the continuity and periodicity of time series for traffic forecasting · Transactions in GIS · 2020 · 385 citations
• A Survey of Traffic Prediction: from Spatio-Temporal Data to Intelligent Transportation · Data Science and Engineering · 2021 · 373 citations
• Real-time crash risk prediction on arterials based on LSTM-CNN · Accident Analysis & Prevention · 2019 · 362 citations

Most recent work

• DADiffNet: Delay-aware diffusion networks with adaptive subgraphs for large scale traffic forecasting · Neural Networks · 2026
• SGA-Seq: Station-Aware Graph Attention Sequence Network for Cellular Traffic Prediction · IEEE Transactions on Network and Service Management · 2026
• Unraveling Spatio-Temporal Foundation Models via the Pipeline Lens: A Comprehensive Review · IEEE Transactions on Knowledge and Data Engineering · 2026
• How to Break It Down for Building It Up? Theory-Guided Graph Decomposition Learning for Spatiotemporal Traffic Prediction · IEEE Transactions on Pattern Analysis and Machine Intelligence · 2026
• Exploiting attention-driven weather-aware multimodal spatio-temporal fusion for urban traffic flow prediction · Future Generation Computer Systems · 2026
• TUM2TWIN: Introducing the large-scale multimodal urban digital twin benchmark dataset · ISPRS Journal of Photogrammetry and Remote Sensing · 2026
• GCMNet: A global context Mamba network for long-term time series forecasting · Pattern Recognition · 2026
• A Survey of Large Language Models for Traffic Forecasting: Methods and Applications · IEEE Transactions on Big Data · 2026
• Graph Neural Networks for Vehicular Social Networks: Trends, Challenges, and Opportunities · IEEE Transactions on Intelligent Transportation Systems · 2026
• Multimodal spatio-temporal fusion: A generalizable GCN-LSTM with attention framework for urban application · Information Fusion · 2026

Related topics in Engineering