Cross-Domain Conditional Diffusion Models for Time Series Imputation

2025-06-14

AI Summary

Cross-domain time series imputation under high missingness (>60%) and dynamic temporal domain shift poses significant challenges: single-domain imputation methods suffer from poor generalizability, while conventional domain adaptation approaches assume complete observations and thus fail to handle distributional uncertainty induced by missing data. To address this, we propose a three-stage framework: (1) frequency-guided imputation, introducing a novel spectral sharing prior to stabilize cross-domain spectral alignment; (2) a dual-branch conditional diffusion model that disentangles domain-invariant representations from domain-specific temporal dependencies; and (3) output-layer selective consistency regularization to mitigate spurious alignment exacerbated by missingness. Evaluated on three real-world datasets, our method substantially outperforms state-of-the-art approaches, demonstrating robust performance particularly under strong temporal shifts and severe missingness.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Diffusion Models for Vision
• Application Domains: Humanities & Computational Social Science

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Cross-domain time series imputation is an underexplored data-centric research task that presents significant challenges, particularly when the target domain suffers from high missing rates and domain shifts in temporal dynamics. Existing time series imputation approaches primarily focus on the single-domain setting, which cannot effectively adapt to a new domain with domain shifts. Meanwhile, conventional domain adaptation techniques struggle with data incompleteness, as they typically assume the data from both source and target domains are fully observed to enable adaptation. For the problem of cross-domain time series imputation, missing values introduce high uncertainty that hinders distribution alignment, making existing adaptation strategies ineffective. Specifically, our proposed solution tackles this problem from three perspectives: (i) Data: We introduce a frequency-based time series interpolation strategy that integrates shared spectral components from both domains while retaining domain-specific temporal structures, constructing informative priors for imputation. (ii) Model: We design a diffusion-based imputation model that effectively learns domain-shared representations and captures domain-specific temporal dependencies with dedicated denoising networks. (iii) Algorithm: We further propose a cross-domain consistency alignment strategy that selectively regularizes output-level domain discrepancies, enabling effective knowledge transfer while preserving domain-specific characteristics. Extensive experiments on three real-world datasets demonstrate the superiority of our proposed approach. Our code implementation is available here.

Problem

• Addressing high missing rates in cross-domain time series imputation
• Overcoming domain shifts in temporal dynamics for imputation
• Handling data incompleteness in cross-domain adaptation tasks

Innovation

• Frequency-based time series interpolation strategy
• Diffusion-based imputation model design
• Cross-domain consistency alignment strategy