Predictive analyses of traffic level in the city of Barcelona: from ARIMA to eXtreme Gradient Boosting

This study delves into the intricate dynamics of urban mobility, a pivotal aspect for policymakers, businesses, and communities alike. By deciphering patterns of movement within a city, stakeholders can craft targeted interventions to mitigate traffic congestion peaks, optimizing both resource alloc...

Full description

Bibliographic Details
Authors: Garcia Climent, Eloi, Calvet Liñán, Laura, Carracedo Garnateo, Patricia, Serrat Piè, Carles|||0000-0002-1504-5354, Miró Martínez, Pau, Peyman, Mohammad
Format: article
Publication Date:2024
Country:España
Institution:Universitat Politècnica de Catalunya (UPC)
Repository:UPCommons. Portal del coneixement obert de la UPC
Language:English
OAI Identifier:oai:upcommons.upc.edu:2117/411523
Online Access:https://hdl.handle.net/2117/411523
https://dx.doi.org/10.3390/app14114432
Access Level:Open access
Keyword:Traffic flow -- Forecasting
Traffic level
eXtreme Gradient Boosting
Forecasting
Mobility
Open data
Circulació -- Previsió
Àrees temàtiques de la UPC::Matemàtiques i estadística::Anàlisi numèrica::Modelització matemàtica
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
Description
Summary:This study delves into the intricate dynamics of urban mobility, a pivotal aspect for policymakers, businesses, and communities alike. By deciphering patterns of movement within a city, stakeholders can craft targeted interventions to mitigate traffic congestion peaks, optimizing both resource allocation and individual travel routes. Focused on Barcelona, Spain, this paper draws on data sourced from the city council’s open data service. Through a blend of exploratory analysis, visualization techniques, and modeling methodologies—including time series analysis and the eXtreme Gradient Boosting (XGBoost) algorithm—the research endeavors to forecast traffic conditions. Additionally, a study of variable importance is carried out, and Shapley Additive Explanations are applied to enhance the interpretability of model outputs. Findings underscore the limitations of traditional forecasting methods in capturing the nuanced spatial and temporal dependencies present in traffic flows, particularly over medium- to long-term horizons. However, the XGBoost model demonstrates robust performance, with the area under ROC curves consistently exceeding 80%, indicating its efficacy in handling non-linear traffic data variables.