Prediction of the traffic incident duration using statistical and machine-learning methods: A systematic literature review

Abstract

This paper aims to present a comprehensive review and analysis to demonstrate the main papers, journals, authors, and trends significantly contributing to the scientific output in predicting the traffic incident duration using statistical and ML-based methods. We analyze new methods as well as data resources and characteristics such as incident time phases, data types, incident types, duration time distribution, available data resources, significant influencing factors, and unobserved heterogeneity and randomness. Also, this paper used the VOSviewer® software to conduct a visualization study of knowledge mapping on the literature of predicting traffic incident duration from 2010 to 2022 based on various databases. The contributions of this paper are three-fold. First, this paper undertakes a comprehensive comparison of previous studies in this field. Second, this paper identifies the key conceptual characteristics of analysis and prediction. Third, this paper explores the expected future trends in predicting the traffic incident duration. Furthermore, a key finding from this paper is that the usage of crowdsourcing, social media, and textual data is rare. Considering that future prediction methods of traffic incident duration will likely be utilized in all subfields related to traffic congestion, focusing on a review that summarizes such studies is a timely topic.