Abstract
Eliminating the random noise is the key issue to be solved to deeply explore the inherent patterns of passenger flow fluctuation and apply accurate prediction. This paper proposes a multi-stage data denoising based short-term section passenger flow patterns extraction and prediction method. Coarse-grained denoising is performed by adaptive soft thresholding of the Empirical Wavelet Transform (EWT) model, and the component attributes are categorized based on the Gramian Angular Summation Field (GASF) images and Determinism (DET) indicator. The periodic inherent patterns of passenger flow fluctuation are analysed by recomposing the deterministic components, and finally fine-grained denoising and trend prediction are achieved by the Gate Recurrent Unit (GRU) algorithm. Numerical experimental results show that the multi-stage denoising model outperforms the noise-balance based time-domain method and accurately identifies periodic fluctuation patterns. Moreover, the proposed EWT-GRU model in this paper significantly improves the accuracy of section passenger flow prediction.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call