Forecasting Method of Urban Rail Transit Ridership at Station-Level on the Basis of Back Propagation Neural Network
Lists
Only you see this
Abstract
Urban rail transit ridership at station-level is an important part of urban rail transit ridership, critical for determining the scale of station and access facilities. Present direct forecasting method of ridership is based on multivariate linear regression with population weighted by distance-decay function. These methods are not able to reflect non-linear relationship between ridership and its predictors. It is also inappropriate to use population in these linear models since it is not uniformly distributed. In this paper, a new variable, population per distance band, is considered and a back propagation neural network model (BPNN) which can reflect non-linear relationship between ridership and its predictors is proposed to forecast ridership with key factors that affect ridership obtained through partial correlation analysis. The performance of the model is compared with three other models, using four MOEs, maximum relative error, smallest relative error, average relative error and mean square root of relative error. Also, a model of contribution rate of population per distance band to ridership is formulated based on the BPNN model when non-population variables are fixed as background variables. Case studies with Japanese data show the effectiveness of our model. The result shows all MOEs of BPNN are much better than the other three models. Contribution rate of population within special distance band to ridership calculated through the model is very close to actual statistical value, and the reason for difference between the two is explained. Ridership can be calculated using results from contribution rate model directly and more quickly by multiplying population within special band and corresponding rate without operating BPNN once more.
BibTeX
@inproceedings{junfang2015forecasting,
author = {LI Jun-fang and Xiafei Ye and Jiaqi Ma},
title = {{Forecasting Method of Urban Rail Transit Ridership at Station-Level on the Basis of Back Propagation Neural Network}},
booktitle = {Transportation},
year = {2015},
}Loading related content...