01663544

Component

Bike-sharing systems are an important part of urban mobility in many cities and are sustainable, environmentally-friendly systems. Since the demand of bikes in stations is still not well studied, this paper introduced an effective approach to quantifying the effect of various features on the prediction of bike counts at each station in the San Francisco Bay Area Bike Share. The Random Forest technique was used to rank the predictors, then guided forward step-wise regression and Bayesian Information Criterion were used to develop and compare bike share prediction models, respectively. The final results demonstrate that the time-of-day, temperature, and humidity level (which has not previously been studied) are significant count predictors. It also shows that weather variables are geographic location dependent. Additionally, findings show that variability of bike counts at some stations is not critical if used as a predictor in the regression.

This paper was sponsored by TRB committee ANF20 Standing Committee on Bicycle Transportation. Alternate title: Quantifying the Effect of Various Features on the Modeling of Bike Counts in a Bikesharing System

18-00282

English

0000-0002-5845-2929

15p

2018

Transportation Research Board 97th Annual Meeting

Digital/other

Figures; Maps; References; Tables

Accessibility; Bicycling; Forecasting; Ridesharing; Traffic counts

San Francisco Bay Area

Operations and Traffic Management; Pedestrians and Bicyclists

Transportation Research Board Annual Meeting 2018 Paper #18-00282

TRIS, TRB, ATRI

Jan 8 2018 10:05AM