Exploring Intercity Trip Patterns of Railway Systems on National Holidays Using Deep Auto-Encoder
Journal
Transportation Research Record
Journal Volume
2674
Journal Issue
5
Pages
662-672
Date Issued
2020
Author(s)
Abstract
Intercity railway system operation on national holidays can be challenging because of possible surging demand. This study proposes an analysis framework to investigate railway system ridership data on national holidays, seeking to attain better understanding of relevant intercity trip patterns, so as to enable enhanced preparation and response before and during national holidays. The ridership data are analyzed in the form of Origin–Destination (O-D) tables and regarded as pictures of N × N pixels, where N is the number of the considered stations/cities in a railway system. The framework primarily adopts a deep auto-encoder to process these pictures to reduce data dimensions and abstracting key features within these pictorial data. Based on the abstracted features, k-means clustering is then conducted to categorize the O-D tables with similar trip patterns into the same group. Further, a discrete outcome model based on logistic regression is developed on the clustering results to enhance the interpretation of the trip pattern in each group and identify the significant holiday-related characteristics and external factors that can affect the trip pattern generation. The ridership data of Taiwan Railways Administration associated with 38 national holidays from January 2014 to August 2018 are analyzed. The analysis results highlight insightful interpretation in relation to clustered trip patterns and relevant trip characterization relative to various national holidays. The proposed framework and developed discrete outcome model are also validated, showing 85% correct assignments of O-D tables to the groups of relevant trip patterns. © National Academy of Sciences: Transportation Research Board 2020.
SDGs
Other Subjects
Abstracting; K-means clustering; Learning systems; Logistic regression; Railroads; Signal encoding; Analysis frameworks; Clustering results; Data dimensions; External factors; Inter-city railway; Model-based OPC; Pictorial data; Railway system; Railroad transportation
Type
journal article