Mobility Modeling and Prediction in Bike-Sharing Systems

Abstract

As an innovative mobility strategy, public bike-sharing has grown dramatically worldwide. Though providing convenient, low-cost and environmental-friendly transportation, the unique features of bike-sharing systems give rise to problems to both users and operators. The primary issue among these problems is the uneven distribution of bicycles caused by the ever-changing usage and (available) supply. This bicycle imbalance issue necessitates efficient bike re-balancing strategies, which depends highly on bicycle mobility modeling and prediction. In this paper, for the first time, we propose a spatio-temporal bicycle mobility model based on historical bike-sharing data, and devise a traffic prediction mechanism on a per-station basis with sub-hour granularity. We extensively evaluated the performance of our design through a oneyear dataset from the world's largest public bike-sharing system (BSS) with more than 2800 stations and over 103 million check in/out records. Evaluation results show an 85 percentile relative error of 0.6 for both check in and check out prediction. We believe this new mobility modeling and prediction approach can advance the bike re-balancing algorithm design and pave the way for the rapid deployment and adoption of bike-sharing systems across the globe.

Spatiotemporal analysis of urban mobility dynamics: a case study of bicycle sharing system

“…Other studied data sets are from Chicago [73,80,98,99], Lyon [63,65,74,100], Boston [12,76,80,101], Barcelona [11,70,102], Hangzhou [15,16,103], Brisbane [61,83], Minneapolis [76,104], Vienna [105,106], Denver [76,84], Pisa [64,107], Dublin [14,108], Minnesota [84], Seville [102], Montreal [109] Helsinki [110], Vancouver [111], Nanjing [112], and Castellon [113]. In addition to BSS data, some of those studies also used weather data as a feature of their analyses.…”

“…Similar to the work on human mobility prediction in subsection 2.1.7 above, numerous prediction methods have been applied to BSS data. They have been investigated for various prediction purposes such as prediction of traffic flows [15,17], the available of bikes and/or vacant docking slots [47,70], number of trips [16], trips duration [73], level of demand [62,85], over-demand prediction [81], pairwise demand prediction [96], number of usage patterns [27], next place prediction [46], cycle lane usage [109], station occupancy [120], the potential destination station and arrival time [73], waiting time for the next available bike [14], and pair of pickup and return stations that close to the origin and destination places of users [73,96].…”

“…Then, using a tuning process to determine the optimal parameters, the authors achieved a better result using RMLSE as a performance metric than the result for RF without the tuning process. In another study, using a similar RF algorithm, Yang et al [16] conducted traffic prediction of bike check-out and check-in using Hangzhou data. Using CDF and RMSLE as performance metrics, they compared their RF algorithm with three baseline predictors, HA, ARMA, and HP-MSI.…”

“…Previous BSS studies have directly predicted the total usage value [15][16][17]. There appears to be scope to make better prediction of system usage, and to investigate what factors and features significantly influence the prediction performance.…”

“…Recently, most BSS research studies have conducted their analysis and prediction at an aggregate level [15][16][17] to observe the global and local trends, for example at city and cluster scale, since almost all BSS public datasets contain information about trips, but these are not linked to individual users. This study uses what we believe is the only public BSS dataset that provides information on (anonymized) individual users.…”