Boundary-enhanced time series data imputation with long-term dependency diffusion models
This work addresses data imputation challenges in fields like healthcare and traffic, offering an incremental improvement over prior diffusion-based approaches.
The paper tackles the problem of disharmonious boundaries and overlooked long-term dependencies in diffusion model-based imputation for multivariate time series data, proposing a DSDI framework that achieves superior performance compared to existing methods.
Data imputation is crucial for addressing challenges posed by missing values in multivariate time series data across various fields, such as healthcare, traffic, and economics, and has garnered significant attention. Among various methods, diffusion model-based approaches show notable performance improvements. However, existing methods often cause disharmonious boundaries between missing and known regions and overlook long-range dependencies in missing data estimation, leading to suboptimal results. To address these issues, we propose a Diffusion-based time Series Data Imputation (DSDI) framework. We develop a weight-reducing injection strategy that incorporates the predicted values of missing points with reducing weights into the reverse diffusion process to mitigate boundary inconsistencies. Further, we introduce a multi-scale S4-based U-Net, which combines hierarchical information from different levels via multi-resolution integration to capture long-term dependencies. Experimental results demonstrate that our model outperforms existing imputation methods.