TSGuard: A Real-Time Framework for Detecting and Imputing Missing Data in Streaming Time Series

2026-10-02

AI Summary

This study addresses the challenges of missing values in streaming sensor data and the lack of real-time capability and domain constraint verification in existing methods. We propose a closed-loop data quality management framework that integrates detection, imputation, validation, and interpretation. The framework employs a lightweight graph-aware temporal imputation model, innovatively incorporating the imputation mechanism into a data quality loop. It combines physical-spatial constraints with fallback estimation to make retention or replacement decisions. Furthermore, the system supports real-time monitoring and interactive validation, ensuring that imputed values strictly adhere to domain specifications. Consequently, this work significantly enhances both the trustworthiness and interpretability of streaming data.

Abstract

Streaming sensor applications routinely suffer from delayed or missing observations caused by faults, communication losses, or environmental interference. Although recent imputation methods exploit temporal and spatial dependencies effectively, most either assume offline access to future observations or prioritize throughput without enforcing domain plausibility. We present TSGuard, a real-time demonstration system for monitoring, validating, and imputing missing values in streaming time series. TSGuard combines a lightweight graph-aware temporal imputation model with constraint-aware validation, fallback estimation, and operator-facing explanations. Rather than treating imputation as an isolated prediction task, TSGuard integrates it into a broader data-quality loop: detect problematic observations, impute missing values, validate estimated against physical and spatial constraints, and either retain the original value as a plausible anomaly or replace it when it violates domain constraints. Using environmental sensing as a motivating setting, the demo enables users to inspect delayed sensors, compare imputers, define constraints, and validate flagged values in real time. The combination of lightweight online spatiotemporal imputation, domain-aware validation, and explicit retain-or-replace decisions is our central contribution, while interactive explanations make these decisions inspectable and actionable. for operators.

Problem

• streaming time series
• missing data imputation
• real-time detection
• constraint-aware validation
• sensor data quality

Innovation

• streaming time series
• real-time imputation
• graph-aware model
• constraint-aware validation
• data quality loop