Caterpillar-SSA is a model-free time series analysis software that identifies, analyzes, and forecasts time series structure using the Singular Spectrum Analysis method. It also supports multivariate analysis.
This software review discusses the advantages and capabilities of a program based on the Caterpillar (SSA) time series analysis method. This technique offers a powerful model-free approach to analyzing time series data that combines the benefits of other methods with simple visual control aids.
The basic algorithm involves transforming one-dimensional time series into a trajectory matrix through a delay procedure. The trajectory matrix then undergoes singular value decomposition, and the original time series is reconstructed based on selected eigenvectors. The output is a natural decomposition of the time series into separate components, such as trends, seasonalities, oscillatory series, or noise components.
This method offers forecasting capabilities and can be expanded to process multidimensional time series and change-point detection. It was developed independently in Russia (St. Petersburg, Moscow) and in the UK and USA under the name of SSA, or Singular Spectrum Analysis.
The new book titled "Analysis of Time Series Structure: SSA and Related Techniques" provides greater depth into the method's capabilities, authored by N. Golyandina, V. Nekrutkin, and A. Zhigljavsky. The program has proven useful in various scientific fields, including meteorology, hydrology, geophysics, climatology, economics, biology, physics, and medicine.
Overall, this software offers a powerful tool for time series analysis and will likely become a base method for statistical software in the future.
The basic algorithm involves transforming one-dimensional time series into a trajectory matrix through a delay procedure. The trajectory matrix then undergoes singular value decomposition, and the original time series is reconstructed based on selected eigenvectors. The output is a natural decomposition of the time series into separate components, such as trends, seasonalities, oscillatory series, or noise components.
This method offers forecasting capabilities and can be expanded to process multidimensional time series and change-point detection. It was developed independently in Russia (St. Petersburg, Moscow) and in the UK and USA under the name of SSA, or Singular Spectrum Analysis.
The new book titled "Analysis of Time Series Structure: SSA and Related Techniques" provides greater depth into the method's capabilities, authored by N. Golyandina, V. Nekrutkin, and A. Zhigljavsky. The program has proven useful in various scientific fields, including meteorology, hydrology, geophysics, climatology, economics, biology, physics, and medicine.
Overall, this software offers a powerful tool for time series analysis and will likely become a base method for statistical software in the future.
What's New
Version 3.40: N/A