|
Features of Caterpillar 3.40
Analysis of one-dimensional time series
- Decomposition of one-dimensional time series into eigentriples (eigenvalues, eigenvectors and principal components)
- Convenient graphical visualization of results for identification of the eigentriples corresponding to trend, periodicities, noise
- Grouping of eigentriples that leads to expansion of the time series into additive components
- Reconstruction of time series components (trend, oscillations, periodicities, noise) by choice of eigentriples
- Residual analysis
|
Forecast of one-dimensional time series
- Approximation (local) of time series by finite-rank series
- Forecast by vector and recurrent methods
- Analyzing the linear recurrent formula used for the recurrent forecast method
- Confidence intervals by empirical and bootstrap methods
- Construction of envelopes (channels)
- Testing the forecast results on validation period
|
Change-point detection for one-dimensional time series
- Change-point detection by comparing the 'Caterpillar-SSA' structures of the base and test time series intervals
- Construction of heterogeneity matrix and detection functions
- Analyzing the found structural changes by moving root and modulus functions
|
Multichannel Analysis/Forecast of time series
- Simultaneous decomposition of several one-dimensional time series into common eigentriples (eigenvalues, eigenvectors and principal components)
- Convenient graphical visualization of results for identification of the eigentriples corresponding to trend, common periodicities, noise
- Grouping of eigentriples that leads to expansion of the time series into additive components
- Reconstruction of the time series components (trend, oscillations, periodicities, noise) by choice of eigentriples
- Approximation (local) of time series by finite-rank series
- Forecast by vector and recurrent methods
- Testing the forecast results on validation period
|
Features of CatMV 1.0
Main features:
- Decomposition of time series with missing values
- Extraction of trend and periodic components
- Approximation of the extracted component by a time series of finite rank with the help of Cadzow iterations method
- Filling in the missing data in the extracted component
- Forecasting the extracted component by setting missing values after non-missing ones
Additional features:
- Navigation through the program stages/results
- Loading data from file
- Simulation of model input data
- Editing the loaded data
- Saving results to a text file (only for registered users)
- Comparison of the reconstructed time series with a specified time series
- Editing graphics options by double click
Features of CatSSA 2.0 (DLL)
- Decomposition of one-dimensional time series into eigentriples (eigenvalues, eigenvectors and principal components) by Basic or Toeplitz SSA
- Reconstruction of time series components (trend, oscillations, periodicities, noise) by choice of eigentriples
- Forecast by the recurrent and vector methods
Features of Caterpillar 1.00
- Decomposition of one-dimensional time series into eigentriples (eigenvalues, eigenvectors and principal components)
- Convenient graphical visualization of results for identification of the eigentriples corresponding to trend, periodicities, noise
- Reconstruction of time series components (trend, oscillations, periodicities, noise) by choice of eigentriples
|
