A Unified Hyperparameter Optimization Pipeline for Transformer-Based Time Series Forecasting Models

TL;DR

The paper addresses the challenge of hyperparameter optimization for transformer-based time series forecasting (TSF) models and the lack of a unified pipeline. It proposes a unified HPO pipeline and benchmarks SOTA models including Autoformer, Crossformer, Non-Stationary Transformer, PatchTST, Mamba, and TimeMixer across ETTh1, Weather, and Electricity datasets, using OptunaSearch (TPE) via Ray Tune and Weights & Biases for visualization. The experimental setup explores a broad search space with 357 trials and analyzes the influence of hyperparameters such as $d ext{-}model$, learning rate, and batch size on performance, while highlighting out-of-memory (OOM) challenges. The work provides practical guidance for practitioners and researchers, with code and results publicly available on GitHub, enhancing reproducibility and extensibility.

Abstract

Transformer-based models for time series forecasting (TSF) have attracted significant attention in recent years due to their effectiveness and versatility. However, these models often require extensive hyperparameter optimization (HPO) to achieve the best possible performance, and a unified pipeline for HPO in transformer-based TSF remains lacking. In this paper, we present one such pipeline and conduct extensive experiments on several state-of-the-art (SOTA) transformer-based TSF models. These experiments are conducted on standard benchmark datasets to evaluate and compare the performance of different models, generating practical insights and examples. Our pipeline is generalizable beyond transformer-based architectures and can be applied to other SOTA models, such as Mamba and TimeMixer, as demonstrated in our experiments. The goal of this work is to provide valuable guidance to both industry practitioners and academic researchers in efficiently identifying optimal hyperparameters suited to their specific domain applications. The code and complete experimental results are available on GitHub.

Figures (6)

• Figure 1: HPO pipeline for Transformer-based forecasting
• Figure 2: Common Parameters in the Transformer
• Figure 3: Training loss and validation loss on each model's best performance case in experiments
• Figure 4: Parallel coordinates plot on Weather dataset: Autoformer without outlier
• Figure 5: Parallel coordinates plot on ETTh1 dataset