About | Contact Us | Register | Login
ProceedingsSeriesJournalsSearchEAI
Intelligent Transport Systems. 8th International Conference, INTSYS 2024, Pisa, Italy, December 5–6, 2024, Revised Selected Papers

Research Article

An AutoML Approach for Bike Demand Forecasting and Redistribution

Cite
BibTeX Plain Text
  • @INPROCEEDINGS{10.1007/978-3-031-86370-7_9,
        author={Dimitris Petratos and Yannis Poulakis and Irene Gimenez Pedralba and Cristina Aragon Garcia and Christos Doulkeridis},
        title={An AutoML Approach for Bike Demand Forecasting and Redistribution},
        proceedings={Intelligent Transport Systems. 8th International Conference, INTSYS 2024, Pisa, Italy, December 5--6, 2024, Revised Selected Papers},
        proceedings_a={INTSYS},
        year={2025},
        month={4},
        keywords={Bike demand forecasting time series forecasting AutoML bike redistribution minimum-cost flow problem},
        doi={10.1007/978-3-031-86370-7_9}
    }
    
  • Dimitris Petratos
    Yannis Poulakis
    Irene Gimenez Pedralba
    Cristina Aragon Garcia
    Christos Doulkeridis
    Year: 2025
    An AutoML Approach for Bike Demand Forecasting and Redistribution
    INTSYS
    Springer
    DOI: 10.1007/978-3-031-86370-7_9
Dimitris Petratos1, Yannis Poulakis1, Irene Gimenez Pedralba, Cristina Aragon Garcia, Christos Doulkeridis1,*
  • 1: Department of Digital Systems
*Contact email: cdoulk@unipi.gr

Abstract

In this paper we introduce a two-staged pipeline to tackle the problem of bike redistribution for bike-sharing systems, using Automated Machine Learning (AutoML) and optimization techniques. Our approach includes the usage of AutoML for time series forecasting in order to estimate the demand for bikes for each station, along with an optimization model to efficiently relocate bikes to maximize user satisfaction. In our study, we used historical data from Barcelona’s public bike-sharing system to predict future demand and then used these predictions together with public data from OpenStreetMap (estimated travel time between stations) in order to solve the Minimum Cost Flow Problem (MCFP) and compute the optimal bike redistribution. We demonstrate promising results in terms of accuracy of demand forecasting and reduction of forecasting time, thus obtaining feasible redistribution strategies and providing an end-to-end framework to the operator.

Keywords
Bike demand forecasting time series forecasting AutoML bike redistribution minimum-cost flow problem
Published
2025-04-03
Appears in
SpringerLink
http://dx.doi.org/10.1007/978-3-031-86370-7_9
Copyright © 2024–2025 ICST
EBSCOProQuestDBLPDOAJPortico
EAI Logo

About EAI

  • Who We Are
  • Leadership
  • Research Areas
  • Partners
  • Media Center

Community

  • Membership
  • Conference
  • Recognition
  • Sponsor Us

Publish with EAI

  • Publishing
  • Journals
  • Proceedings
  • Books
  • EUDL