
201Accelerating Removal of Variation
of highest to lowest so that the PC loading vectors are in order of signi-
cance.
18
For the example, as p
3
captures the most variability of the data,
p
3
is arranged to be the rst eigenvector, and its eigenvalue is arranged
accordingly.
The number of PC loading vectors that should be retained to form the
feature vector is dependent on the variance captured by the PC loading
vectors. The percentage of the variance captured by every PC (EV) is sim-
ply calculated by using the following equation:
∑
=
100
1
EV
l
l
i
i
i
i
m
where l is the variance or eigenvalue, i is the selected PC loading vector,
and m is the number of ...