Predicting Uber Demand of NYC with Wavenet

Chen, L., Thakuriah, V. and Ampountolas, K. (2019) Predicting Uber Demand of NYC with Wavenet. In: SMART ACCESSIBILITY 2019, Athens, Greece, 24-28 Feb 2019, ISBN 9781612086910

Full text not currently available from Enlighten.


Uber demand prediction is at the core of intelligent transportation systems when developing a smart city. However, exploiting uber real time data to facilitate the demand prediction is a thorny problem since user demand usually unevenly distributed over time and space. We develop a Wavenet-based model to predict Uber demand on an hourly basis. In this paper, we present a multi-level Wavenet framework which is a one-dimensional convolutional neural network that includes two sub-networks which encode the source series and decode the predicting series, respectively. The two sub-networks are combined by stacking the decoder on top of the encoder, which in turn, preserves the temporal patterns of the time series. Experiments on large-scale real Uber demand dataset of NYC demonstrate that our model is highly competitive to the existing ones.

Item Type:Conference Proceedings
Glasgow Author(s) Enlighten ID:Thakuriah, Professor Piyushimita and Chen, Dr Long and Ampountolas, Dr Konstantinos
Authors: Chen, L., Thakuriah, V., and Ampountolas, K.
Subjects:T Technology > T Technology (General)
T Technology > TA Engineering (General). Civil engineering (General)
College/School:College of Science and Engineering > School of Engineering > Infrastructure and Environment
College of Social Sciences > School of Social and Political Sciences
Related URLs:

University Staff: Request a correction | Enlighten Editors: Update this record

Project CodeAward NoProject NamePrincipal InvestigatorFunder's NameFunder RefLead Dept
301865RCUK (UKRI) Industrial Fellowship 2017 - UBDCJinhyun HongEconomic and Social Research Council (ESRC)ES/S001875/1S&PS - Urban Studies