An online learning approach to eliminate Bus Bunching in real-time

[Display omitted] •We proposed a data driven method to predict Bus Bunching (BB) in real-time.•Firstly, the bus headways are predicted combining Offline and Online Regression.•Then, a BB likelihood is computed for each stop based on such predictions.•Finally, a corrective action is selected and depl...

Full description

Saved in:
Bibliographic Details
Published in:Applied soft computing 2016-10, Vol.47 (C), p.460-482
Main Authors: Moreira-Matias, Luís, Cats, Oded, Gama, João, Mendes-Moreira, João, de Sousa, Jorge Freire
Format: Article
Language:English
Subjects:
Bus Bunching
Online learning
Stochastic Gradient Descent
Travel Time Prediction
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:[Display omitted] •We proposed a data driven method to predict Bus Bunching (BB) in real-time.•Firstly, the bus headways are predicted combining Offline and Online Regression.•Then, a BB likelihood is computed for each stop based on such predictions.•Finally, a corrective action is selected and deployed using these likelihoods.•This methodology is validated using one-year data of 18 real-world bus routes. Recent advances in telecommunications created new opportunities for monitoring public transport operations in real-time. This paper presents an automatic control framework to mitigate the Bus Bunching phenomenon in real-time. The framework depicts a powerful combination of distinct Machine Learning principles and methods to extract valuable information from raw location-based data. State-of-the-art tools and methodologies such as Regression Analysis, Probabilistic Reasoning and Perceptron's learning with Stochastic Gradient Descent constitute building blocks of this predictive methodology. The prediction's output is then used to select and deploy a corrective action to automatically prevent Bus Bunching. The performance of the proposed method is evaluated using data collected from 18 bus routes in Porto, Portugal over a period of one year. Simulation results demonstrate that the proposed method can potentially reduce bunching by 68% and decrease average passenger waiting times by 4.5%, without prolonging in-vehicle times. The proposed system could be embedded in a decision support system to improve control room operations.
ISSN:1568-4946
1872-9681
DOI:10.1016/j.asoc.2016.06.031