Evaluation Framework for Train Delays Prediction Models
Tiong, Kah Yong LU ; Ma, Zhenliang and Palmqvist, Carl-William LU
(2023)
- Abstract
- Real-time train delay prediction is critical for ensur-
ing the safety of train operations and providing quality service
to passengers. The increasing demand for railway transportation
has led to the need for punctuality and the availability of advance
train arrival time information to ensure a satisfactory travel
experience. Data-driven train prediction models have undergone
significant advancements, and there is a growing emphasis
on hybrid approaches that combine multiple data-driven and
simulation methods. However, there is still a lack of clarity
regarding the evaluation procedure for assessing the quality of
these models. To address this, we propose a three-stage evaluation
framework that... (More) - Real-time train delay prediction is critical for ensur-
ing the safety of train operations and providing quality service
to passengers. The increasing demand for railway transportation
has led to the need for punctuality and the availability of advance
train arrival time information to ensure a satisfactory travel
experience. Data-driven train prediction models have undergone
significant advancements, and there is a growing emphasis
on hybrid approaches that combine multiple data-driven and
simulation methods. However, there is still a lack of clarity
regarding the evaluation procedure for assessing the quality of
these models. To address this, we propose a three-stage evaluation
framework that disaggregates the process of evaluating train
delay prediction algorithms into problem identification, evalu-
ation, and monitoring clusters. This study presents a critical
discussion clarifying the interactions and logical flow between
the stages of the proposed evaluation framework to provide
guidance for evaluating and selecting prediction models. The
main contributions of this paper are the introduction of a novel
approach for evaluating real-time train delay prediction at three
levels of data aggregation, framing, and discussing the necessary
steps for the proposed prediction evaluation framework. The
proposed framework aims to equip railway researchers and
practitioners with the necessary tools to bridge the gap between
the development and implementation of train delay prediction
models in the real world. By enabling critical evaluations of data-
driven prediction models before deploying them, the framework
can help to enhance the accuracy and effectiveness of train
delay prediction, thereby improving the quality of service for
passengers and ensuring the safety of train operations. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/e5f315a6-e91b-4ebe-9a16-42e52ebd9888
- author
- Tiong, Kah Yong
LU
; Ma, Zhenliang
and Palmqvist, Carl-William
LU
- organization
- publishing date
- 2023
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- in press
- subject
- host publication
- 7th International Conference on Transportation Information and Safety (ICTIS 2023)
- publisher
- IEEE - Institute of Electrical and Electronics Engineers Inc.
- language
- English
- LU publication?
- yes
- id
- e5f315a6-e91b-4ebe-9a16-42e52ebd9888
- date added to LUP
- 2023-07-16 16:40:33
- date last changed
- 2023-08-02 11:31:18
@inproceedings{e5f315a6-e91b-4ebe-9a16-42e52ebd9888,
abstract = {{Real-time train delay prediction is critical for ensur-<br/>ing the safety of train operations and providing quality service<br/>to passengers. The increasing demand for railway transportation<br/>has led to the need for punctuality and the availability of advance<br/>train arrival time information to ensure a satisfactory travel<br/>experience. Data-driven train prediction models have undergone<br/>significant advancements, and there is a growing emphasis<br/>on hybrid approaches that combine multiple data-driven and<br/>simulation methods. However, there is still a lack of clarity<br/>regarding the evaluation procedure for assessing the quality of<br/>these models. To address this, we propose a three-stage evaluation<br/>framework that disaggregates the process of evaluating train<br/>delay prediction algorithms into problem identification, evalu-<br/>ation, and monitoring clusters. This study presents a critical<br/>discussion clarifying the interactions and logical flow between<br/>the stages of the proposed evaluation framework to provide<br/>guidance for evaluating and selecting prediction models. The<br/>main contributions of this paper are the introduction of a novel<br/>approach for evaluating real-time train delay prediction at three<br/>levels of data aggregation, framing, and discussing the necessary<br/>steps for the proposed prediction evaluation framework. The<br/>proposed framework aims to equip railway researchers and<br/>practitioners with the necessary tools to bridge the gap between<br/>the development and implementation of train delay prediction<br/>models in the real world. By enabling critical evaluations of data-<br/>driven prediction models before deploying them, the framework<br/>can help to enhance the accuracy and effectiveness of train<br/>delay prediction, thereby improving the quality of service for<br/>passengers and ensuring the safety of train operations.}},
author = {{Tiong, Kah Yong and Ma, Zhenliang and Palmqvist, Carl-William}},
booktitle = {{7th International Conference on Transportation Information and Safety (ICTIS 2023)}},
language = {{eng}},
publisher = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}},
title = {{Evaluation Framework for Train Delays Prediction Models}},
year = {{2023}},
}