Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Jinliang Deng, Feiyang Ye, Du Yin, Xuan Song, Ivor W. Tsang, Hui Xiong

TL;DR

This study demonstrates, through both analytical and empirical evidence, that decomposition is key to containing excessive model inflation while achieving uniformly superior and robust results across various datasets.

Abstract

Long-term time series forecasting (LTSF) represents a critical frontier in time series analysis, characterized by extensive input sequences, as opposed to the shorter spans typical of traditional approaches. While longer sequences inherently offer richer information for enhanced predictive precision, prevailing studies often respond by escalating model complexity. These intricate models can inflate into millions of parameters, resulting in prohibitive parameter scales. Our study demonstrates, through both analytical and empirical evidence, that decomposition is key to containing excessive model inflation while achieving uniformly superior and robust results across various datasets. Remarkably, by tailoring decomposition to the intrinsic dynamics of time series data, our proposed model outperforms existing benchmarks, using over 99 \% fewer parameters than the majority of competing methods. Through this work, we aim to unleash the power of a restricted set of parameters by capitalizing on domain characteristics--a timely reminder that in the realm of LTSF, bigger is not invariably better.

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

TL;DR

This study demonstrates, through both analytical and empirical evidence, that decomposition is key to containing excessive model inflation while achieving uniformly superior and robust results across various datasets.

Abstract

Long-term time series forecasting (LTSF) represents a critical frontier in time series analysis, characterized by extensive input sequences, as opposed to the shorter spans typical of traditional approaches. While longer sequences inherently offer richer information for enhanced predictive precision, prevailing studies often respond by escalating model complexity. These intricate models can inflate into millions of parameters, resulting in prohibitive parameter scales. Our study demonstrates, through both analytical and empirical evidence, that decomposition is key to containing excessive model inflation while achieving uniformly superior and robust results across various datasets. Remarkably, by tailoring decomposition to the intrinsic dynamics of time series data, our proposed model outperforms existing benchmarks, using over 99 \% fewer parameters than the majority of competing methods. Through this work, we aim to unleash the power of a restricted set of parameters by capitalizing on domain characteristics--a timely reminder that in the realm of LTSF, bigger is not invariably better.
Paper Structure (38 sections, 19 equations, 9 figures, 3 tables)

This paper contains 38 sections, 19 equations, 9 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Selective Structured Components-based Neural Network
  4. Temporal Component
  5. Long-Term Component
  6. Seasonal Component
  7. Short-Term Component
  8. Spatial Component
  9. Component Fusion
  10. Comparison Between Decomposition and Patching
  11. Capability Analysis
  12. Parsimony Analysis
  13. Evaluations
  14. Experiment Setting
  15. Datasets
  16. ...and 23 more sections

Figures (9)

  • Figure 1: An overview of the SSCNN. The grids are used to exemplify the selection maps $\mathcal{I}^*$ and $\mathcal{E}^*$ as defined in the main text, with $T_\text{in}$, $T_\text{out}$ and $N$ instantiated as 4, 4 and 3, respectively.
  • Figure 2: Examination of parameter scale and computation scale against the forward window size and the backward window size on the ECL dataset.
  • Figure 3: Impacts of backward window size.
  • Figure 4: Sensitivity analysis of hyper-parameters on the ECL and Traffic datasets.
  • Figure 5: Performance comparison with various component in ECL and Traffic dataset.
  • ...and 4 more figures

Theorems & Definitions (2)

  • Definition 1: Conditional Correlation
  • Definition 2: Conditional Auto-correlation