I am working on Singular Spectrum Analysis. I had done the Embedding Step & Decomposition Step step. I stuck at Grouping Step. Here i want to know how to do the grouping. At what basis we should take the numeric group index & define a group of window indices.
I am working with Pavia dataset (one pixel vector).
Dear Shrish Bajpai,
1) Forms of eigenvectors repeats the form of the component, which you want to extract. You can use this this as a promp for grouping
2) You can use so-called w-correlations between elementary components reconstructed on the base of elementary grouping (1 index = 1 group). Then the components with high correlation should be included into one group.
3) You can use frequency approach, grouping components with close frequency ranges. This approach is used in auto-grouping implemented in the R-package Rssa https://cran.r-project.org/package=Rssa by the function grouping.auto.pgram.
- Badges
- Science topic
- Similar topics
- Vectorization