Citation

Myhrmann MS, Mabit SE. Transp. Res. A Policy Pract. 2023; 176: e103783.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.tra.2023.103783

PMID

unavailable

Abstract

Cycling can reduce greenhouse gas emissions and air pollution and increase public health. Hence, policymakers in cities worldwide seek to improve bicycle mode shares. Efforts to increase the bicycle's mode share involve many measures, one of them being the improvement of cycling safety often requiring an analysis of the factors surrounding accidents. However, meaningful analysis of cycling safety requires accurate bicycle flow data that are generally sparse or only available at the aggregate level. Therefore, safety engineers often rely on aggregated variables or calibration factors that fail to account for variations in the cycling traffic relevant to policymaking. This paper illustrates how machine learning can support policy analysis by delivering detailed bicycle flow predictions. The illustration applies a Deep Learning approach, the Long Short-Term Memory Mixture Density Network (LSTMMDN), to estimate hourly bicycle flow in Copenhagen, conditional on weather, temporal and road conditions at the segment level. The method addresses some shortcomings in the calibration factor method resulting in 66-77% more accurate bicycle traffic estimates. To quantify the impact of more accurate bicycle traffic estimates in cycling safety analysis, we test the effect of different flow estimates in a bicycle crash risk model, i.e. the models are identical except for the exposure variables. One model is estimated using the LSTMMDN estimates, one using the calibration-based estimates, and one using yearly mean traffic estimates. The results show that investing in more advanced methods for obtaining bicycle volume estimates can improve the quality of safety analyses and other performance measures.

Language: en

Keywords

Aggregation bias; Bicycle flow estimation; Deep Learning; Long Short-Term Memory; Mixture Density Network