Revisiting the Seasonal Trend Decomposition for Enhanced Time Series Forecasting

TL;DR

The paper addresses multivariate time series forecasting by decomposing signals into trend and seasonal components using a simple moving-average approach. It models the trend with reversible instance normalization (RevIN) around a small MLP and processes the seasonal part with backbones stripped of normalization, complemented by two efficient dual-MLP models (RMSM and RMM) that operate in $O(L)$ time. The authors demonstrate consistent improvements—approximately a 10% average reduction in MSE across four baselines and notable gains on USGS hydrological data—while reducing computational overhead compared to Transformer-based backbones. The work highlights the practical value of simple decomposition strategies for real-world forecasting tasks and provides open-source code for replication and extension.

Abstract

Time series forecasting presents significant challenges in real-world applications across various domains. Building upon the decomposition of the time series, we enhance the architecture of machine learning models for better multivariate time series forecasting. To achieve this, we focus on the trend and seasonal components individually and investigate solutions to predict them with less errors. Recognizing that reversible instance normalization is effective only for the trend component, we take a different approach with the seasonal component by directly applying backbone models without any normalization or scaling procedures. Through these strategies, we successfully reduce error values of the existing state-of-the-art models and finally introduce dual-MLP models as more computationally efficient solutions. Furthermore, our approach consistently yields positive results with around 10% MSE average reduction across four state-of-the-art baselines on the benchmark datasets. We also evaluate our approach on a hydrological dataset extracted from the United States Geological Survey (USGS) river stations, where our models achieve significant improvements while maintaining linear time complexity, demonstrating real-world effectiveness. The source code is available at https://github.com/Sanjeev97/Time-Series-Decomposition

Figures (1)

• Figure 1: Illustration of our approach with moving average and showing iTransformer (as backbone) with two other alternatives: Shift-MLP and MLP.