Missing traffic data imputation with a conditional diffusion framework

Jie Li

doi:10.1016/j.eswa.2025.129140

Missing traffic data imputation with a conditional diffusion framework

2025-07-24

0

PRE

AI

DOI：10.1016/j.eswa.2025.129140

Original

Save

Abstract

En 中文

• Address data imputation via VAE ensuring spatiotemporal consistency and multi-scale refinement of temporal patterns. • Conditional feature module captures spatiotemporal patterns in input data to enhance noise estimation accuracy. • VDM: conditional diffusion framework leveraging spatiotemporal features for noise estimation and reverse-process data imputation. • Tested on 5 real-world traffic datasets, our model demonstrates superior imputation performance over baseline methods.

Keywords:

VAE

conditional diffusion model

spatiotemporal consistency

data imputation

traffic data

Missing traffic data imputation with a conditional diffusion framework

Abstract

Journal

Organization

Cited Papers