Categorizing the Basic Using Patterns of Bicycle Sharing System and Investigating the Effects of Location Factors: A Non-Negative Matrix Factorization Approach

Abstract

The use of bicycle sharing systems (BSS) has generated a large amount of data for transport and urban studies This study aims to categorized BSS using patterns and to evaluate the relationship between location factors and BSS use Non-negative matrix factorization (NMF) was employed which can disintegrate a large amount of data into several more comprehensible components The relationships between BSS uses and surrounding location factors were investigated using linear regression Data were extracted using the BSS database from Taipei City Taiwan First the BSS trips were categorized by renting durations and the distances between origin and destination (OD) to differentiate borrow-return behavior patterns Then the NMF was performed to decompose the overall temporal distribution of trips in BSS stations into several sets of peak-time using patterns Each pattern was assumed to represent the urban activities generated from some specific location factors Finally linear regression models were built to test the relationship between the BSS using patterns and location factors The outcomes of this study show that the overall temporal distribution of trips in a BSS station is aggregated from the trips of several BSS user groups and the NMF can extract the latent patterns within the overall data In addition we used a novel microscopic approach to categorize each trip by its attributes and found that the type of borrow-return behavior plays a significant role on analyzing BSS using patterns This research contributes to a better understanding of BSS travel behavior supporting decision making in bicycle policy and urban planning

Date of Award	2020
Original language	English
Supervisor	Tzu-Chang Lee (Supervisor)