Travel mode choice prediction: developing new techniques to prioritize variables and interpret black-box machine learning techniques
Document Type
Journal Article
Publication Date
2025
Subject Area
ridership - demand, ridership - forecasting, ridership - mode choice, ridership - modelling, planning - methods
Keywords
Travel mode choice, classification algorithms, feature analysis, gray relational analysis
Abstract
Travel Mode Choice (TMC) prediction is vital for forecasting travel demand and transportation planning. To be helpful for those purposes, one needs to know with high accuracy what influences choices and how. For accuracy, Machine Learning (ML) classification techniques often produce results with higher accuracy than traditional methods. However, many ML techniques are black-box tools, making them less useful for planning. To this end, two new approaches were proposed to interpret the results of ML techniques and investigate the influence of different variables on TMC. The results suggested that ensemble learning techniques outperform other prediction methods. Adding accessibility, geographic, and land-use variables to the conventional TMC prediction models could improve their performance. The most important parameters for TMC were found to be: trip distance, availability of a transit pass and availability of a driver’s license. Their respective influences on the different modes are demonstrated using the novel method mentioned above.
Rights
Permission to publish the abstract has been given by Taylor&Francis, copyright remains with them.
Recommended Citation
Naseri, H., Waygood, E. O. D., Patterson, Z., Alousi-Jones, M., & Wang, B. (2025). Travel mode choice prediction: developing new techniques to prioritize variables and interpret black-box machine learning techniques. Transportation Planning and Technology, 48(3), 582-605.
