Task-oriented Time Series Imputation Evaluation via Generalized Representers

TL;DR

This paper introduces a novel evaluation approach for time series imputation that focuses on how well imputed data supports downstream tasks like forecasting, using neural network models to estimate the impact without retraining.

Contribution

It proposes a task-oriented evaluation method that assesses imputation quality based on downstream task performance, integrating multiple strategies for optimal results.

Findings

Effective estimation of imputation impact on downstream tasks

Improved selection of imputation strategies for specific tasks

Enhanced understanding of imputation effects without retraining models

Abstract

Time series analysis is widely used in many fields such as power energy, economics, and transportation, including different tasks such as forecasting, anomaly detection, classification, etc. Missing values are widely observed in these tasks, and often leading to unpredictable negative effects on existing methods, hindering their further application. In response to this situation, existing time series imputation methods mainly focus on restoring sequences based on their data characteristics, while ignoring the performance of the restored sequences in downstream tasks. Considering different requirements of downstream tasks (e.g., forecasting), this paper proposes an efficient downstream task-oriented time series imputation evaluation approach. By combining time series imputation with neural network models used for downstream tasks, the gain of different imputation strategies on downstream tasks is estimated without retraining, and the most favorable imputation value for downstream tasks is given by combining different imputation strategies according to the estimated gain.