A method for filling traffic data based on feature-based combination prediction model

Abstract Data imputation is a critical step in data processing, directly influencing the accuracy of subsequent research. However, due to the temporal nature of ride-hailing trajectory data, traditional imputation methods often struggle to adequately consider spatiotemporal characteristics, leading...

Full description

Saved in:
Bibliographic Details
Main Authors: Haicheng Xiao, Xueyan Shen, Jianglin Li, Xiujian Yang
Format: Article
Language:English
Published: Nature Portfolio 2025-03-01
Series:Scientific Reports
Subjects:
Online Access:https://doi.org/10.1038/s41598-025-92547-y
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract Data imputation is a critical step in data processing, directly influencing the accuracy of subsequent research. However, due to the temporal nature of ride-hailing trajectory data, traditional imputation methods often struggle to adequately consider spatiotemporal characteristics, leading to limitations in both convergence speed and accuracy. To address this issue, this study employs a prediction-based approach to enhance imputation accuracy. Given the limited feature parameters in trajectory data, traditional prediction models often fail to comprehensively capture data characteristics. Therefore, this study proposes a feature generation model based on LightGBM-GRU, combined with a SARIMA-GRU prediction model, to more thoroughly capture and enrich the data characteristics. This approach effectively imputes missing data, thereby laying a solid foundation for subsequent research.
ISSN:2045-2322