Towards Better Long-range Time Series Forecasting using Generative Forecasting

TL;DR

This paper introduces Generative Forecasting (GenF), a novel strategy that uses generative models to produce synthetic data for improved long-range time series forecasting, outperforming existing methods.

Contribution

The paper proposes a new forecasting strategy called GenF that balances bias and variance, utilizing a novel GAN-based generator, a transformer predictor, and an information-theoretic clustering algorithm.

Findings

GenF achieves 5-11% lower mean absolute error compared to benchmarks.

GenF reduces model parameters by 15-50% relative to existing methods.

Experimental results on five datasets validate the effectiveness of GenF.

Abstract

Long-range time series forecasting is usually based on one of two existing forecasting strategies: Direct Forecasting and Iterative Forecasting, where the former provides low bias, high variance forecasts and the latter leads to low variance, high bias forecasts. In this paper, we propose a new forecasting strategy called Generative Forecasting (GenF), which generates synthetic data for the next few time steps and then makes long-range forecasts based on generated and observed data. We theoretically prove that GenF is able to better balance the forecasting variance and bias, leading to a much smaller forecasting error. We implement GenF via three components: (i) a novel conditional Wasserstein Generative Adversarial Network (GAN) based generator for synthetic time series data generation, called CWGAN-TS. (ii) a transformer based predictor, which makes long-range predictions using both generated and observed data. (iii) an information theoretic clustering algorithm to improve the training of both the CWGAN-TS and the transformer based predictor. The experimental results on five public datasets demonstrate that GenF significantly outperforms a diverse range of state-of-the-art benchmarks and classical approaches. Specifically, we find a 5% - 11% improvement in predictive performance (mean absolute error) while having a 15% - 50% reduction in parameters compared to the benchmarks. Lastly, we conduct an ablation study to further explore and demonstrate the effectiveness of the components comprising GenF.