PCA algorithm in WEKA

0
votes

I'm using a data-set with 22 attributes in WEKA and after applying the PrincipalComponents attribute evaluator I get 8 ranked attributes with 0.95 variance covered. As far as I know, the sum of the new values of the attributes should be equal to 0.95, but for some reason my ranked attribute variance doesn't add up to. What's the issue with it? All of the attributes are numeric, except for the class attribute which is nominal.

Here it is - the output I get in WEKA after PCA: Ranked attributes

Any help is appreciated, thank you.

1
attributeswekapcavariance

1 Answers

0
votes

The variances being totaled by yourself are scores derived from cumulative variances.

Please refer

  1. Answer by Mark Hall:

    http://weka.8497.n7.nabble.com/PCA-Output-td3579.html

And

  1. Q&A on PCA variances add to 100%

    https://stats.stackexchange.com/questions/32901/do-components-of-pca-really-represent-percentage-of-variance-can-they-sum-to-mo