Add PCA class

[paulk-asert]

As per discussion on the forum:
https://forum.orekit.org/t/potential-principal-component-analysis-component/2175

[maisonobe]

Thanks for the contribution!

[axkr]

Can we have a more general PCA implementation like in Mathematica, which also allows a Method->"Correlation" option, additional to the default Method->"Covariance"?

• https://reference.wolfram.com/language/ref/PrincipalComponents.html

[paulk-asert]

This is already supported. You need the constructor with additional options for scaling (correlation):
https://github.com/Hipparchus-Math/hipparchus/blob/master/hipparchus-stat/src/main/java/org/hipparchus/stat/projection/PCA.java#L76-L83

Using new PCA(2) gives the equivalent* output in the Mathematica basic example.

Using new PCA(3, true, true) gives the equivalent* output as the example shown under "Options".

* The absolute values are the same but the sign differs.