Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LMS-AutoTSF: Learnable Multi-Scale Decomposition and Integrated Autocorrelation for Time Series Forecasting

Ibrahim Delibasoglu, Sanjay Chakraborty, Fredrik Heintz

TL;DR

LMS-AutoTSF tackles multivariate time-series forecasting by introducing learnable, multi-scale frequency-domain decomposition and integrated autocorrelation. Each scale learns separate low-pass (trend) and high-pass (seasonality) components via trainable FFT-based filters, while lagged differences inject temporal dependencies into a lightweight, fully-connected encoder framework. Across diverse benchmarks, it achieves state-of-the-art accuracy with notably reduced computation, demonstrating strong performance on long- and short-term horizons and high-dimensional data. The approach offers a practical, scalable solution for real-world forecasting with public code availability.

Abstract

Time series forecasting is an important challenge with significant applications in areas such as weather prediction, stock market analysis, scientific simulations and industrial process analysis. In this work, we introduce LMS-AutoTSF, a novel time series forecasting architecture that incorporates autocorrelation while leveraging dual encoders operating at multiple scales. Unlike models that rely on predefined trend and seasonal components, LMS-AutoTSF employs two separate encoders per scale: one focusing on low-pass filtering to capture trends and the other utilizing high-pass filtering to model seasonal variations. These filters are learnable, allowing the model to dynamically adapt and isolate trend and seasonal components directly in the frequency domain. A key innovation in our approach is the integration of autocorrelation, achieved by computing lagged differences in time steps, which enables the model to capture dependencies across time more effectively. Each encoder processes the input through fully connected layers to handle temporal and channel interactions. By combining frequency-domain filtering, autocorrelation-based temporal modeling, and channel-wise transformations, LMS-AutoTSF not only accurately captures long-term dependencies and fine-grained patterns but also operates more efficiently compared to other state-of-the-art methods. Its lightweight design ensures faster processing while maintaining high precision in forecasting across diverse time horizons. The source code is publicly available at \url{http://github.com/mribrahim/LMS-TSF}

LMS-AutoTSF: Learnable Multi-Scale Decomposition and Integrated Autocorrelation for Time Series Forecasting

TL;DR

LMS-AutoTSF tackles multivariate time-series forecasting by introducing learnable, multi-scale frequency-domain decomposition and integrated autocorrelation. Each scale learns separate low-pass (trend) and high-pass (seasonality) components via trainable FFT-based filters, while lagged differences inject temporal dependencies into a lightweight, fully-connected encoder framework. Across diverse benchmarks, it achieves state-of-the-art accuracy with notably reduced computation, demonstrating strong performance on long- and short-term horizons and high-dimensional data. The approach offers a practical, scalable solution for real-world forecasting with public code availability.

Abstract

Time series forecasting is an important challenge with significant applications in areas such as weather prediction, stock market analysis, scientific simulations and industrial process analysis. In this work, we introduce LMS-AutoTSF, a novel time series forecasting architecture that incorporates autocorrelation while leveraging dual encoders operating at multiple scales. Unlike models that rely on predefined trend and seasonal components, LMS-AutoTSF employs two separate encoders per scale: one focusing on low-pass filtering to capture trends and the other utilizing high-pass filtering to model seasonal variations. These filters are learnable, allowing the model to dynamically adapt and isolate trend and seasonal components directly in the frequency domain. A key innovation in our approach is the integration of autocorrelation, achieved by computing lagged differences in time steps, which enables the model to capture dependencies across time more effectively. Each encoder processes the input through fully connected layers to handle temporal and channel interactions. By combining frequency-domain filtering, autocorrelation-based temporal modeling, and channel-wise transformations, LMS-AutoTSF not only accurately captures long-term dependencies and fine-grained patterns but also operates more efficiently compared to other state-of-the-art methods. Its lightweight design ensures faster processing while maintaining high precision in forecasting across diverse time horizons. The source code is publicly available at \url{http://github.com/mribrahim/LMS-TSF}

Paper Structure

This paper contains 8 sections, 9 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Problem Statement
  4. Methodology
  5. Experiments
  6. Datasets
  7. Forecasting Results
  8. Conclusion

Figures (5)

  • Figure 1: (a) Overall architecture of LMS-AutoTS. (b) Encoder module operations for trend and seasonal components in Scale-2
  • Figure 2: Comparison of models efficiency with datasets vs. avg. MSE vs. avg. MAE
  • Figure 3: Visualization of prediction results on ETTh2 dataset
  • Figure 4: Visualization of prediction results on Electricity dataset
  • Figure 5: Visualization of prediction results on PEMS03 dataset