Malek Sarhani , Abtin Nourmohammadzadeh , Stefan Voß , Mohammed EL Amrani
{"title":"Predicting and analyzing ferry transit delays using open data and machine learning","authors":"Malek Sarhani , Abtin Nourmohammadzadeh , Stefan Voß , Mohammed EL Amrani","doi":"10.1016/j.jpubtr.2025.100124","DOIUrl":null,"url":null,"abstract":"<div><div>The utilization of public transport data has evolved rapidly in recent decades. Ferries, with their unique characteristics and sensitivity to weather conditions, pose significant challenges for delay prediction. Given their pivotal role in the transportation systems of numerous cities, accurately predicting ferry delays is crucial for synchronizing transit services.</div><div>This paper demonstrates the value of open data for improving ferry delay predictions through machine learning, focusing on two case studies. Our approach leverages General Transit Feed Specification (GTFS) data, ridership and vessel information, and hourly weather data, combined with SHAP explainable artificial intelligence analysis to assess key delay determinants. While support vector regression and deep neural networks showed high accuracy in individual case studies, gradient boosting consistently offered the best balance between prediction accuracy and computational efficiency. Moreover, SHAP analysis reveals that operational and temporal features – such as stop sequence, trip start time, headway, and vehicle label – are the dominant drivers of delays, with weather-related factors exerting only a modest influence.</div></div>","PeriodicalId":47173,"journal":{"name":"Journal of Public Transportation","volume":"27 ","pages":"Article 100124"},"PeriodicalIF":2.0000,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Public Transportation","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1077291X25000098","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"TRANSPORTATION","Score":null,"Total":0}
引用次数: 0
Abstract
The utilization of public transport data has evolved rapidly in recent decades. Ferries, with their unique characteristics and sensitivity to weather conditions, pose significant challenges for delay prediction. Given their pivotal role in the transportation systems of numerous cities, accurately predicting ferry delays is crucial for synchronizing transit services.
This paper demonstrates the value of open data for improving ferry delay predictions through machine learning, focusing on two case studies. Our approach leverages General Transit Feed Specification (GTFS) data, ridership and vessel information, and hourly weather data, combined with SHAP explainable artificial intelligence analysis to assess key delay determinants. While support vector regression and deep neural networks showed high accuracy in individual case studies, gradient boosting consistently offered the best balance between prediction accuracy and computational efficiency. Moreover, SHAP analysis reveals that operational and temporal features – such as stop sequence, trip start time, headway, and vehicle label – are the dominant drivers of delays, with weather-related factors exerting only a modest influence.
期刊介绍:
The Journal of Public Transportation, affiliated with the Center for Urban Transportation Research, is an international peer-reviewed open access journal focused on various forms of public transportation. It publishes original research from diverse academic disciplines, including engineering, economics, planning, and policy, emphasizing innovative solutions to transportation challenges. Content covers mobility services available to the general public, such as line-based services and shared fleets, offering insights beneficial to passengers, agencies, service providers, and communities.