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

TL;DR

This paper introduces PIR, a model-agnostic post-hoc framework that identifies and revises biased forecasts in time series data, significantly improving accuracy and reliability across various models and real-world datasets.

Contribution

The paper presents PIR, a novel post-hoc revision method that enhances time series forecasts by addressing instance-level biases through contextual information.

Findings

PIR reduces forecasting errors on real-world datasets.

PIR improves reliability of various forecasting models.

PIR effectively mitigates instance-specific biases.

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.