Impute On-Demand: Adaptive Correlated Time Series Imputation for Changing Environments

2026-07-26

AI Summary

This work addresses the challenge of efficiently and adaptively imputing missing values in IoT time series data, which often arise from sensor failures or dynamic changes in sensor subsets—particularly under resource-constrained conditions where existing methods struggle. The authors propose AdaCTSi, a lightweight framework that decouples spatial and temporal features via a one-shot temporal convolutional network coupled with a learnable spatiotemporal index table. AdaCTSi incorporates sparse spatial attention and a correlation-weighted sensor selection mechanism to enable adaptive imputation for arbitrary incomplete sensor subsets. Notably, it supports dynamic adaptation to varying sensor configurations and computational budgets within a single model, overcoming the limitations of static architectures. Evaluated on five benchmark datasets, AdaCTSi reduces the average MAE by 33.1% compared to the strongest baseline and demonstrates efficient deployability on resource-limited devices such as microcontrollers.

Abstract

Internet of Things (IoT) applications generate vast amounts of Correlated Time Series (CTS) data that often contain missing values and require imputation. Existing methods emphasize accuracy but often lack adaptability to changing IoT environments: they are vulnerable to sensor failures, cannot selectively impute only incomplete sensors, and use static architectures that do not adapt to resource availability. To address these limitations, we propose AdaCTSi, an adaptive CTS imputer for changing environments. AdaCTSi combines a One-shot Temporal Convolutional Network with a Learned Time-Sensor Index Table to extract and decouple complex spatio-temporal features into sensor-wise embeddings, enabling adaptation to varying sensor subsets. Sparse Spatial Attention efficiently extracts dynamic spatial correlations, while Correlation-Weighted Sensor Selection selects informative sensors to provide sufficient spatial context. Experiments with twelve baseline methods, three adaptability scenarios, and five benchmark datasets covering traffic, air quality, and trajectory data show that AdaCTSi reduces MAE by an average of 33.1% relative to the strongest baseline on each dataset. A single trained model supports sensor-subset and resource-adaptive inference, and its modest memory footprint enables deployment on commodity computing devices, including MCUs.

Problem

• Correlated Time Series
• Missing Data Imputation
• Adaptability
• IoT Environments
• Sensor Failure

Innovation

• Adaptive Imputation
• Correlated Time Series
• Sparse Spatial Attention
• Sensor Selection
• Resource-Efficient Deployment