Optimizing Time Series Forecasting Architectures: A Hierarchical Neural Architecture Search Approach

TL;DR

This paper introduces a hierarchical neural architecture search method tailored for time series forecasting, enabling the automatic design of efficient and high-performing models by combining various forecasting modules.

Contribution

It presents a novel hierarchical search space that effectively integrates multiple forecasting modules, improving architecture design for time series tasks.

Findings

Successfully searches for lightweight high-performing architectures

Effective across different long-term forecasting tasks

Outperforms some existing methods in accuracy and efficiency

Abstract

The rapid development of time series forecasting research has brought many deep learning-based modules in this field. However, despite the increasing amount of new forecasting architectures, it is still unclear if we have leveraged the full potential of these existing modules within a properly designed architecture. In this work, we propose a novel hierarchical neural architecture search approach for time series forecasting tasks. With the design of a hierarchical search space, we incorporate many architecture types designed for forecasting tasks and allow for the efficient combination of different forecasting architecture modules. Results on long-term-time-series-forecasting tasks show that our approach can search for lightweight high-performing forecasting architectures across different forecasting tasks.