ForecastGAN: A Decomposition-Based Adversarial Framework for Multi-Horizon Time Series Forecasting

TL;DR

ForecastGAN is a new adversarial framework that combines decomposition and neural network selection to improve multi-horizon time series forecasting, especially excelling in short-term predictions and handling categorical data.

Contribution

It introduces a decomposition-based adversarial framework with modules for trend extraction, model selection, and adversarial training, effectively integrating categorical features and outperforming existing models.

Findings

Outperforms state-of-the-art transformer models in short-term forecasting

Maintains competitive performance across long-term horizons

Works effectively on diverse multivariate datasets

Abstract

Time series forecasting is essential across domains from finance to supply chain management. This paper introduces ForecastGAN, a novel decomposition based adversarial framework addressing limitations in existing approaches for multi-horizon predictions. Although transformer models excel in long-term forecasting, they often underperform in short-term scenarios and typically ignore categorical features. ForecastGAN operates through three integrated modules: a Decomposition Module that extracts seasonality and trend components; a Model Selection Module that identifies optimal neural network configurations based on forecasting horizon; and an Adversarial Training Module that enhances prediction robustness through Conditional Generative Adversarial Network training. Unlike conventional approaches, ForecastGAN effectively integrates both numerical and categorical features. We validate our framework on eleven benchmark multivariate time series datasets that span various forecasting horizons. The results show that ForecastGAN consistently outperforms state-of-the-art transformer models for short-term forecasting while remaining competitive for long-term horizons. This research establishes a more generalizable approach to time series forecasting that adapts to specific contexts while maintaining strong performance across diverse data characteristics without extensive hyperparameter tuning.