Abstract:This paper introduces SparseTSF, a novel, extremely lightweight model for Long-term Time Series Forecasting (LTSF), designed to address the challenges of modeling complex temporal dependencies over extended horizons with minimal computational resources. At the heart of SparseTSF lies the Cross-Period Sparse Forecasting technique, which simplifies the forecasting task by decoupling the periodicity and trend in time series data. This technique involves downsampling the original sequences to focus on cross-period trend prediction, effectively extracting periodic features while minimizing the model's complexity and parameter count. Based on this technique, the SparseTSF model uses fewer than *1k* parameters to achieve competitive or superior performance compared to state-of-the-art models. Furthermore, SparseTSF showcases remarkable generalization capabilities, making it well-suited for scenarios with limited computational resources, small samples, or low-quality data. The code is publicly available at this repository: <a class="link-external link-https" href="https://github.com/lss-1138/SparseTSF" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

### The Problem the Paper Attempts to Solve This paper aims to address several key challenges in Long-term Time Series Forecasting (LTSF): 1. **Complex Temporal Dependencies**: Long-term time series forecasting requires models to capture complex temporal dependencies over long time windows. This often demands models with high complexity and a large number of parameters. 2. **Computational Resource Constraints**: Existing high-performance models (such as Transformer) perform well in terms of prediction accuracy, but their large parameter sizes make them difficult to apply in scenarios with limited computational resources. 3. **Separation of Data Seasonality and Trend**: Time series data typically exhibit significant seasonality and trend characteristics. Effectively separating and utilizing these features to simplify the forecasting task is an important research direction. To this end, the paper proposes the **SparseTSF** model, which introduces **Cross-Period Sparse Forecasting** technology to compress the model parameters to below 1k while maintaining high prediction performance. Specifically, SparseTSF addresses the above issues through the following methods: - **Downsampling**: The original time series is downsampled according to a fixed period to generate multiple subsequences. - **Sparse Forecasting**: Each downsampled subsequence is predicted, transforming the original time series forecasting task into a cross-period trend forecasting task. - **Aggregation and Upsampling**: After prediction, the subsequences are re-aggregated and upsampled to restore the complete forecast sequence. Through this approach, SparseTSF can effectively extract periodic features, focus on predicting trend changes, and significantly reduce model complexity and parameter size, making it suitable for scenarios with limited computational resources.

SparseTSF: Modeling Long-term Time Series Forecasting with 1k Parameters

Effective LSTMs with Seasonal-Trend Decomposition and Adaptive Learning and Niching-Based Backtracking Search Algorithm for Time Series Forecasting

The Bigger the Better? Rethinking the Effective Model Scale in Long-term Time Series Forecasting

Parsimony or Capability? Decomposition Delivers Both in Long-term Time Series Forecasting

MixLinear: Extreme Low Resource Multivariate Time Series Forecasting with 0.1K Parameters

FoundTS: Comprehensive and Unified Benchmarking of Foundation Models for Time Series Forecasting

Less Is More: Fast Multivariate Time Series Forecasting with Light Sampling-oriented MLP Structures

How Much Can Time-related Features Enhance Time Series Forecasting?

A lightweight multi-layer perceptron for efficient multivariate time series forecasting

Improving Long-Term Electricity Time Series Forecasting in Smart Grid with a Three-Stage Channel-Temporal Approach

Deep Time Series Forecasting Models: A Comprehensive Survey

Modeling Temporal Dependencies within the Target for Long-Term Time Series Forecasting

FlexTSF: A Universal Forecasting Model for Time Series with Variable Regularities

Adapt to Small-Scale and Long-Term Time Series Forecasting with Enhanced Multidimensional Correlation

CycleNet: Enhancing Time Series Forecasting through Modeling Periodic Patterns

SimpleTS: an Efficient and Universal Model Selection Framework for Time Series Forecasting

Time Series Forecasting (TSF) Using Various Deep Learning Models

Multivariate Time Series Modeling and Forecasting with Parallelized Convolution and Decomposed Sparse-Transformer

CTFNet: Long-Sequence Time-Series Forecasting Based on Convolution and Time–Frequency Analysis

Attractor Memory for Long-Term Time Series Forecasting: A Chaos Perspective

MvTS-library: an Open Library for Deep Multivariate Time Series Forecasting