Improving Time Series Forecasting via Instance-aware Post-hoc Revision

2025-05-29

AI Summary

In time series forecasting, instance-level biases—such as distributional shifts, missing values, and long-tailed patterns—cause localized prediction inaccuracies that remain undetected by aggregate evaluation metrics. To address this, we propose PIR, a model-agnostic post-hoc framework that introduces a novel paradigm: *post-prediction accuracy estimation → bias-aware instance identification → context-aware dynamic correction*. PIR requires no model retraining; instead, it employs a confidence-driven, instance-aware mechanism that jointly leverages local sliding-window features and global statistical patterns, augmented by covariate-guided residual correction. Evaluated across multiple real-world benchmarks, PIR consistently enhances state-of-the-art models—including Informer, Autoformer, and PatchTST—reducing MAE and MSE by 5.2%–12.7% on average. Notably, it significantly improves robustness and reliability for long-tailed and anomalous instances, demonstrating strong generalizability and practical utility.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: State Estimation
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Time series forecasting plays a vital role in various real-world applications and has attracted significant attention in recent decades. While recent methods have achieved remarkable accuracy by incorporating advanced inductive biases and training strategies, we observe that instance-level variations remain a significant challenge. These variations--stemming from distribution shifts, missing data, and long-tail patterns--often lead to suboptimal forecasts for specific instances, even when overall performance appears strong. To address this issue, we propose a model-agnostic framework, PIR, designed to enhance forecasting performance through Post-forecasting Identification and Revision. Specifically, PIR first identifies biased forecasting instances by estimating their accuracy. Based on this, the framework revises the forecasts using contextual information, including covariates and historical time series, from both local and global perspectives in a post-processing fashion. Extensive experiments on real-world datasets with mainstream forecasting models demonstrate that PIR effectively mitigates instance-level errors and significantly improves forecasting reliability.

Problem

• Addressing instance-level variations in time series forecasting
• Mitigating suboptimal forecasts due to distribution shifts and missing data
• Improving forecasting reliability via post-hoc identification and revision

Innovation

• Model-agnostic framework for post-forecasting revision
• Identifies biased instances via accuracy estimation
• Revises forecasts using local and global context