Road Traffic Forecasting: Recent Advances and New Challenges

Laña, Ibai; Del Ser, Javier; Vélez, Manuel; Vlahogianni, Eleni I. · 2018 · OpenAlex-citations

DOI: 10.1109/mits.2018.2806634

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper provides a comprehensive review of road traffic forecasting, analyzing the evolution of methodologies from early time-series approaches to modern data-driven techniques. Motivated by the critical need for accurate traffic predictions to mitigate congestion and support Intelligent Transportation Systems (ITS), the authors examine over 40 years of research. The study aims to synthesize prior surveys, identify technical advancements, and highlight unresolved challenges, particularly in the context of Big Data and machine learning. The authors conducted a systematic examination of literature, categorizing studies based on prediction methods, horizons, scales, data sources, and exogenous factors. They analyzed the shift from parametric models, such as autoregressive integrated moving average (ARIMA) and Kalman filters, to non-parametric machine learning approaches like neural networks and support vector machines. The review also assessed the impact of increased data availability, including floating car data (FCD), camera feeds, and social media, on model performance and complexity. Key findings indicate a significant transition toward data-driven forecasting, with a growing emphasis on network-wide and urban arterial predictions, which were previously underrepresented. While short-term prediction remains dominant, recent works demonstrate that long-term forecasts are increasingly viable using large datasets and spatio-temporal correlations. The paper identifies several persistent challenges: the lack of universal model selection criteria, difficulties in comparing heterogeneous models due to disparate performance metrics, and the limited integration of exogenous factors like weather and incidents. Additionally, while data fusion techniques are emerging, most models still rely on single-source data, primarily traffic loops. The significance of this work lies in its identification of specific research niches for future development. The authors conclude that the field must move beyond isolated point predictions to robust network-wide models and develop standardized benchmark datasets for fair model comparison. They emphasize the need for hybrid models that combine spatio-temporal complexity with exogenous inputs to handle the stochastic nature of traffic. By mapping the current state of the art, the paper guides researchers toward addressing gaps in generalizability, model selection, and the effective utilization of diverse data sources for more accurate and actionable traffic forecasts.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• anticipation