Citation

Zhang C, He J, Liu Z, Xing L, Wang Y. PLoS One 2019; 14(10): e0223973.

Affiliation

Smart Transportation Applications and Research Laboratory, Department of Civil and Environmental Engineering, University of Washington, Seattle, WA, United States of America.

Copyright

(Copyright © 2019, Public Library of Science)

DOI

10.1371/journal.pone.0223973

PMID

31618244

Abstract

In order to address the time pattern problems in free-floating car sharing, in this paper, the authors offer a comprehensive time-series method based on deep learning theory. According to car2go booking record data in Seattle area. Firstly, influence of time and location on the free-floating car-sharing usage pattern is analyzed, which reveals an apparent doublet pattern for time and dependence usage amount on population. Then, on the basis of the long-short-term memory recurrent neural network (LSTM-RNN), hourly variation in short-term traffic characteristics including travel demand and travel distance are modeled. The results were also compared with other different statistical models, such as support vector regression (SVR), Autoregressive Integrated Moving Average model (ARIMA), single and second exponential smoothing. It showed that (LSTM-RNN) shows better performance in terms of statistical analysis and tendency precision based on limited data sample.

Language: en