To See Far, Look Close: Evolutionary Forecasting for Long-term Time Series

TL;DR

This paper introduces Evolutionary Forecasting (EF), a new paradigm for long-term time series forecasting that outperforms traditional direct forecasting by mitigating optimization issues and enabling models trained on short horizons to excel at long-term predictions.

Contribution

The paper proposes EF as a unified generative framework, demonstrating that DF is a special case, and shows EF's superior performance and stability over existing methods.

Findings

EF models outperform direct forecasting ensembles on benchmarks.

EF exhibits robust asymptotic stability in extreme extrapolation.

EF enables long-term forecasting with models trained on short horizons.

Abstract

The prevailing Direct Forecasting (DF) paradigm dominates Long-term Time Series Forecasting (LTSF) by forcing models to predict the entire future horizon in a single forward pass. While efficient, this rigid coupling of output and evaluation horizons necessitates computationally prohibitive re-training for every target horizon. In this work, we uncover a counter-intuitive optimization anomaly: models trained on short horizons-when coupled with our proposed Evolutionary Forecasting (EF) paradigm-significantly outperform those trained directly on long horizons. We attribute this success to the mitigation of a fundamental optimization pathology inherent in DF, where conflicting gradients from distant futures cripple the learning of local dynamics. We establish EF as a unified generative framework, proving that DF is merely a degenerate special case of EF. Extensive experiments demonstrate that a singular EF model surpasses task-specific DF ensembles across standard benchmarks and exhibits robust asymptotic stability in extreme extrapolation. This work propels a paradigm shift in LTSF: moving from passive Static Mapping to autonomous Evolutionary Reasoning.