Post: Conference paper: Toward in-depth analysis of train punctuality data

Conference paper: Toward in-depth analysis of train punctuality data

Punctuality (on-time performance, schedule adherence) is one of the most important success factors for a railway traffic system. Despite the importance of punctuality, there seems to be a lack of broad understanding when it comes to the formation of punctuality. Within a railway traffic system, delays concatenate easily, that is, a single delay is likely to cause many other delays, so-called secondary delays. To date, most of the studies related to delays have examined only stations and station-like systems. However, especially within a single-tracked infrastructure – as in Finland – a notable portion of secondary delays takes place outside stations. Thus, the examination of delay concatenation should be done by considering the whole network.

This paper first describes the current practices and challenges related to the analysis of train punctuality data for Finnish railways. Railway organizations in Finland record a lot of punctuality-related data, but with the current data and methods, one is able to allocate primary delays only to their causes. Secondary delays are not analyzed. Considering this, it is also impossible to identify which of the primary delays are the most critical ones. Hence, the research question of this paper is: How can we analyze train punctuality data more efficiently and systematically in order to gain an understanding about the most critical delay concatenation phenomena?

With actual motion data from Finnish railways, we prove that at least the most explicit delay chains can be identified and mapped, and this information can be used, for example, for allocating delays to their real causes, and developing a more robust timetable. We thus argue that this kind of examination can be made and, more importantly, should be made. We also suggest a data mining method called sequence analysis to automate this process. Sequence analysis, with the other data mining techniques, would provide a significant improvement over traditional statistical techniques when analyzing the train punctuality data.

Authors: Jouni Paavilainen & Riikka Salkonen
Title: Toward in-depth analysis of train punctuality data
Conference: WCTR 2010, the 12th World Conference on Transport Research, Lisbon, 11–15 July 2010
Paper ID: 02839
Publication: General Proceedings of the 12th World Conference on Transport Research Society, ISBN 978-989-96986-0-4