View in EDS

Deep-learning-based model for prediction of crowding in a public transit system

Saved in:
Bibliographic Details
Title: Deep-learning-based model for prediction of crowding in a public transit system
Authors: Arpit Shrivastava, Nishtha Rawat, Amit Agarwal
Source: Public Transport. 16:449-484
Publisher Information: Springer Science and Business Media LLC, 2024.
Publication Year: 2024
Subject Terms: ridership - behaviour, Crowding, Affinity propagation, planning - methods, place - asia, operations - crowding, planning - signage/information, Occupancy, Public transport, LSTM, place - urban, technology - ticketing systems
Description: Crowding in public transport is one of the reasons that nudges road users to shift from public transport to private modes of transport. To provide the passengers with a facility to plan their trips as per the dynamic crowding levels, this work proposes a framework for a passenger information system (PIS), in which the transit choices are differentiated with respect to crowding levels on the transit routes at different times of the day. A granular crowding prediction model is developed and integrated with PIS. In this, firstly, the transit segment relation (TSR) is constituted and used to make clusters based on the ridership index. Further, a time-series model is trained for each cluster using boarding TSR. A case study of Bhubaneswar, India, is presented, and three months of ticketing data are used to demonstrate the performance of the proposed prediction model. The prediction model is integrated into the PIS to expedite various route choices.
Document Type: Article
Language: English
ISSN: 1613-7159
1866-749X
DOI: 10.1007/s12469-024-00360-z
Rights: Springer Nature TDM
Accession Number: edsair.doi.dedup.....43192efc4e37abf70e97cb46b41b3941
Database: OpenAIRE
Description
Abstract:Crowding in public transport is one of the reasons that nudges road users to shift from public transport to private modes of transport. To provide the passengers with a facility to plan their trips as per the dynamic crowding levels, this work proposes a framework for a passenger information system (PIS), in which the transit choices are differentiated with respect to crowding levels on the transit routes at different times of the day. A granular crowding prediction model is developed and integrated with PIS. In this, firstly, the transit segment relation (TSR) is constituted and used to make clusters based on the ridership index. Further, a time-series model is trained for each cluster using boarding TSR. A case study of Bhubaneswar, India, is presented, and three months of ticketing data are used to demonstrate the performance of the proposed prediction model. The prediction model is integrated into the PIS to expedite various route choices.
ISSN:16137159
1866749X
DOI:10.1007/s12469-024-00360-z