Open AccessA prediction model of buses passenger flow based on neural networks
Author(s) -
Shuyu Zhang,
Zikang Liu,
Fengting Shen,
Shumeng Wang,
Xuelin Yang
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1656/1/012002
Subject(s) - principal component analysis , artificial neural network , curse of dimensionality , computer science , flow (mathematics) , set (abstract data type) , mean squared prediction error , term (time) , basis (linear algebra) , data set , test data , data mining , artificial intelligence , machine learning , mathematics , physics , geometry , quantum mechanics , programming language
With the popularization of smart buses, the ways to obtain buses operation information are becoming increasingly diversified, and passenger flow prediction based on big data information has also emerged. In this work, Principal Component Analysis (PCA) and error Back Propagation (BP) neural networks were combined to propose a prediction model of short-term buses passenger flow on the basis of PCA-BP neural networks. Firstly, the PCA method was adopted to reduce the dimensionality of the indices of buses passenger flow and improve the input nodes of the BP networks. Secondly, the fully trained network was employed to predict the buses passenger flow. More specifically, the passenger flow data set pair of the 17th bus in Guiyang on November 29, 2019 was used to carry out the PCA-BP neural networks model test. The test results indicated that the predicted passenger flow of each time period was very close, and the passenger flow prediction error in most time periods was very small. Besides, the relative error could achieve a satisfactory fitting effect, which was able to provide a reliable basis for bus dispatch. We found that the proposed PCA-BP neural networks model had high prediction accuracy and prediction performance. Therefore, it is expected to be used as a model for final short-term passenger flow prediction.