Service Analytics

6.Principal Component Analysis

Optimal Basis

  • we want to capture as much variance in the dataset as possible with the first new axis.
  • Then we capture the remaining variance with another new (orthogonal) axis, and so forth . . . . The assumption is that after a while, not much variance will be left.
     
     
    Optimizing variance is equivalent to minimizing the sum of squares.

Diskussion