Complex seasonality in bus urban mobility: assessment of individual and ensemble prediction models
Document Type
Journal Article
Publication Date
2025
Subject Area
place - urban, mode - bus, planning - methods, ridership - perceptions, technology - geographic information systems
Keywords
Total travel time, seasonality, time series, forecast, ensemble, hybrid
Abstract
A bus’s total travel time and its variability are crucial variables that affect user perception, as well as the competitiveness of the service. Total travel time often exhibits multiple seasonal patterns, which invalidates conventional forecasting methods. In this paper, we address the problem of forecasting seasonal data from GPS bus observations using conventional statistical and artificial intelligence models to discuss the handling of multiple seasonal components in urban mobility data. The goal is to shed light on the advantages and disadvantages of several methods selected according to a parsimony gradient, where both simple alternatives (e.g. historical averages or seasonally naïve models) and complex models, including linear and nonlinear ensembles of conventional models, are assessed. The results show that under limited input training data and short-term prediction horizons (one week ahead), simple conventional models perform better than more sophisticated models (e.g. SARIMA or ARNN), whereas ensemble approaches outperform conventional methods.
Rights
Permission to publish the abstract has been given by Taylor&Francis, copyright remains with them.
Recommended Citation
de Zea-Graells, D., Grifoll, M., Estrada, M., & Mension, J. (2025). Complex seasonality in bus urban mobility: assessment of individual and ensemble prediction models. Transportmetrica B: Transport Dynamics, 13(1), 2490510.
