Citation

Zhu S, Zhu F. Transp. Res. A Policy Pract. 2019; 129: 217-231.

Copyright

(Copyright © 2019, Elsevier Publishing)

DOI

10.1016/j.tra.2019.08.009

PMID

unavailable

Abstract

The cycling comfort level of different cycling infrastructure can strongly influence the comfort perception of cyclists and their route choices. In this paper, the cycling comfort index (CCI) is used to measure the cycling comfort level on cycling infrastructure and describe different cycle track characteristics. An Instrumented Probe Bicycle (IPB), which is equipped with a video camera and a set of sensors including GPS receiver, accelerometer, etc., is employed to collect data while being ridden by cyclist in Singapore. An automatic video processing technique using convolutional neural network (CNN) is applied, such that no direct field measurement is required and the data collection process is less time-consuming. Video-based survey is carried out to capture the correlation between CCI and the comfort perception of cyclists. The extreme gradient boosting (XGBoost) method is employed to build the CCI model dependent on various explanatory variables and survey participants' ratings. The results show that the overall accuracy of the XGBoost method is 11% higher than the ordered Probit model commonly used in literature.

Language: en

Keywords

Cycling comfort index; Instrumented Probe Bicycle; Probit model; XGBoost model