RSSA

Rssa is an R package that implements Singular Spectrum Analysis (SSA) and related methods for time series and image data. It provides a framework to decompose a sequence into interpretable components such as trend, periodicities, and noise, using the construction of a trajectory (Hankel) matrix and subsequent singular value decomposition. The package supports both univariate SSA and extensions for multivariate data.

Key features include the creation of trajectory matrices with a user-specified window length, the decomposition of the matrix into elementary components, and the grouping of these components to form reconstructed series. It also offers forecasting and backcasting capabilities through recurrent and vector SSA approaches, enabling short- to intermediate-term predictions based on the identified structure. In addition to time series, Rssa includes tools for two-dimensional SSA (2D-SSA) suitable for image and surface analysis, enabling denoising and feature extraction in visual data.

Rssa is designed for researchers and practitioners working with time series who seek a nonparametric, data-driven method to separate stable patterns from noise. The package supports multichannel analyses, allowing simultaneous treatment of multiple related series via multivariate SSA (MSSA). It provides diagnostic plots and utilities to assess singular values, component time series, and reconstructed components, facilitating interpretation and validation of results.

The software is open-source and available on CRAN and GitHub, with ongoing development to expand methods, improve performance, and enhance documentation. Rssa is used across disciplines such as economics, meteorology, environmental science, and engineering for exploratory data analysis, signal extraction, and forecasting.