[Paper Review] A Comprehensive Survey on Traffic Prediction.

Xueyan Yin, Genze Wu+4 more|arXiv (Cornell University)|Apr 18, 2020

Traffic Prediction and Management Techniques52 citations

TL;DR

This paper presents a comprehensive survey of traffic prediction methods, categorizing existing approaches, evaluating state-of-the-art techniques on public datasets, and benchmarking performance for traffic speed and demand prediction. It provides a taxonomy, reviews applications, compiles key datasets, and identifies future research directions in intelligent transportation systems.

ABSTRACT

Traffic prediction plays an essential role in intelligent transportation system. Accurate traffic prediction can assist route planing, guide vehicle dispatching, and mitigate traffic congestion. This problem is challenging due to the complicated and dynamic spatio-temporal dependencies between different regions in the road network. Recently, a significant amount of research efforts have been devoted to this area, greatly advancing traffic prediction abilities. The purpose of this paper is to provide a comprehensive survey for traffic prediction. Specifically, we first summarize the existing traffic prediction methods, and give a taxonomy of them. Second, we list the common applications of traffic prediction and the state-of-the-art in these applications. Third, we collect and organize widely used public datasets in the existing literature. Furthermore, we give an evaluation by conducting extensive experiments to compare the performance of methods related to traffic demand and speed prediction respectively on two datasets. Finally, we discuss potential future directions.

Motivation & Objective

• To provide a systematic taxonomy of existing traffic prediction methods.
• To review the state-of-the-art applications of traffic prediction in intelligent transportation systems.
• To collect and organize widely used public datasets for traffic prediction research.
• To evaluate and compare the performance of traffic prediction models on speed and demand prediction tasks.
• To identify open challenges and suggest future research directions.

Proposed method

• The paper conducts a structured literature review to classify traffic prediction methods into distinct categories based on their underlying techniques.
• It evaluates state-of-the-art models using extensive experiments on two widely used public datasets for speed and demand prediction.
• The authors organize and summarize commonly used public datasets, including their characteristics and usage in prior research.
• A comparative analysis is performed on models for both traffic speed and demand prediction, focusing on performance metrics and generalization.
• The survey includes a discussion of methodological trends, limitations, and potential improvements in current approaches.
• Future research directions are identified based on gaps in current methodologies and datasets.

Experimental results

Research questions

• RQ1 What are the main categories and taxonomies of traffic prediction methods in the literature?
• RQ2 How do state-of-the-art models perform on public datasets for traffic speed and demand prediction?
• RQ3 Which public datasets are most widely used in traffic prediction research, and what are their key features?
• RQ4 What are the key performance differences between models for speed prediction versus demand prediction?
• RQ5 What are the major open challenges and future research directions in traffic prediction?

Key findings

• The survey identifies multiple methodological categories in traffic prediction, including statistical, machine learning, and deep learning-based approaches.
• Performance benchmarks show that deep learning models, particularly those leveraging graph neural networks and attention mechanisms, outperform traditional methods on both speed and demand prediction tasks.
• The paper compiles a comprehensive list of public datasets, highlighting their spatial and temporal coverage, data types, and usage in prior studies.
• Experimental evaluation reveals consistent performance advantages for graph-based models in capturing spatial dependencies across road networks.
• The study identifies data scarcity, model generalization, and real-time inference as key challenges limiting further progress.
• Future research should focus on explainable AI, multimodal data integration, and improved generalization across diverse urban environments.