Skip to contentScholarGate
LibraryBookshelfDeskReview StudioAssistant
Sign in
On this page
IntuitionHow it worksWhen to use itStrengths & limitationsCommon pitfallsApplicationsFrequently asked🔒 Read the full methodSourcesRelated methods
Cite this pageSpotted an issue on this page? Report or suggest a fix →
Home›Time Series›Singular Spectrum Analysis
Process / pipelineMatrix decomposition and reconstruction

Singular Spectrum Analysis

Also known as: SSA, SVD-based decomposition

Singular Spectrum Analysis (SSA) is a nonparametric method for time-series decomposition and forecasting based on singular value decomposition (SVD) of a time-lagged embedding matrix. Introduced by Broomhead and King (1986) and developed further by Vautard, Yiou, and Ghil (1992), SSA decomposes time series into trend, oscillatory, and noise components without assuming any underlying model. It is particularly effective for short, noisy non-stationary signals where parametric approaches fail.

ScholarGate
  1. Process / pipeline
  2. v1
  3. 3 Sources
  4. PUBLISHED
Cite this page →
Tools & resources
Download slides
Learn & explore

Read the full method

Members only

Sign in with a free account to read this section.

Sign in

Method map

The neighbourhood of related methods — select a node to explore.

Singular Spectrum Analysis
Independent Component An…Kernel PCASingular Value Decomposi…

When to use it

Apply SSA when you need to separate trend, oscillations, and noise in short, noisy time series without assuming a parametric model. It is ideal for climate data, economic indicators, seismic signals, and any domain with non-stationary behavior. SSA shines when signal-to-noise ratio is low or when the data has multiple overlapping components.

Strengths & limitations

Strengths
  • Nonparametric approach—does not require specification of model order or frequency components
  • Works well with short time series and noisy data where classical methods struggle
  • Provides simultaneous trend extraction and noise filtering
  • Interpretable components with clear physical or business meaning
  • Invertible and reversible—reconstructed components exactly sum to the original signal
Limitations
  • Requires careful choice of window length L—sensitivity analysis is needed to ensure robustness
  • Grouping singular components into signal and noise is subjective; no automatic algorithm exists
  • For very long time series, the embedding matrix becomes large, increasing memory and computational cost
  • The method assumes stationarity of patterns within the window—unsuitable for highly non-stationary signals with abrupt regime changes

Frequently asked

How do I choose the optimal window length L?

L should be long enough to capture important patterns (typically L > 2 × dominant period) but short enough to have sufficient degrees of freedom (L < N/2). Use cumulative variance plots or cross-validation to assess sensitivity and guide the choice.

How do I separate signal from noise in the singular values?

The singular value spectrum often has a clear 'knee' where signal values drop off sharply and noise begins. Visual inspection of the scree plot combined with physical interpretation of eigenvectors helps identify the cutoff. More formal methods include MDL (minimum description length) and AIC.

Can SSA forecast into the future?

Yes, the Recurrent SSA (R-SSA) and Vector SSA (VSSA) variants can be used for forecasting. They use the reconstructed trend and oscillatory components to project forward. However, SSA itself is primarily a decomposition tool; forecasting requires additional modeling.

What is the relationship between SSA and PCA?

SSA is essentially PCA applied to a time-lagged embedding matrix. The difference is conceptual: PCA treats observations as independent, while SSA exploits the temporal structure of the data through the embedding matrix.

Can SSA handle multivariate time series?

Yes, multivariate SSA (MSSA) extends the method to handle several time series simultaneously. It constructs a block Hankel matrix with columns from multiple series and applies the same SVD and reconstruction steps.

Sources

  1. Broomhead, D. S., & King, G. P. (1986). Extracting qualitative dynamics from experimental data. Physica D: Nonlinear Phenomena, 20(2–3), 217–236. DOI: 10.1016/0167-2789(86)90031-X ↗
  2. Vautard, R., Yiou, P., & Ghil, M. (1992). Singular-spectrum analysis: A toolkit for short, noisy chaotic signals. Physica D: Nonlinear Phenomena, 58(1–4), 95–126. DOI: 10.1016/0167-2789(92)90103-T ↗
  3. Golyandina, N., Nekrutkin, V., & Zhigljavsky, A. (2001). Analysis of Time Series Structure: SSA and Related Techniques. Chapman and Hall/CRC. link ↗

How to cite this page

ScholarGate. (2026, June 3). Singular Spectrum Analysis. ScholarGate. https://scholargate.app/en/time-series/singular-spectrum-analysis

Related methods

Independent Component AnalysisKernel PCASingular Value Decomposition

Which method?

Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.

  • Independent Component AnalysisMachine learning↔ compare
  • Kernel PCAMachine learning↔ compare
  • Singular Value DecompositionNumerical Methods↔ compare
Compare side by side →

Similar methods

Maximum Covariance AnalysisTime Series Kalman FilterSTL DecompositionRecurrence Quantification AnalysisEmpirical Orthogonal TeleconnectionState Space ModelEmpirical Wavelet TransformNon-negative Matrix Factorization

Related reference concepts

Principal Component AnalysisDimension ReductionMatrix Decompositions in StatisticsCanonical Correlation AnalysisDimensionality ReductionMultidimensional Scaling

Spotted an issue on this page? Report or suggest a fix →

ScholarGate — Singular Spectrum Analysis (Singular Spectrum Analysis). Retrieved 2026-07-21 from https://scholargate.app/en/time-series/singular-spectrum-analysis · Dataset: https://doi.org/10.5281/zenodo.20539026
Quick facts
Originator
David Broomhead
Subfamily
Matrix decomposition and reconstruction
Year
1986
Type
Dimension reduction and trend extraction
Related methods
Independent Component AnalysisKernel PCASingular Value Decomposition
ScholarGate

A content-first reference library for research methods — what each one is, how it works, and where it comes from.

Open data (CC-BY)

Explore

  • Library
  • Search the library…
  • Browse by field
  • Fields
  • Journey
  • Compare
  • Which method?

Reference

  • Subjects
  • Atlas
  • Glossary
  • Methodology
  • Philosophy

Your tools

  • Bookshelf
  • Desk
  • Chat

Company

  • About
  • Pricing
  • Contact
  • Suggest a method

Entries are compiled from published sources for reference. Verifying the accuracy and suitability of any information for your own use remains your responsibility.

© 2026 ScholarGate · A research-method reference library
  • Privacy
  • Cookies
  • Terms
  • Delete account