Question

In Principal Component Analysis (PCA), what property must the resulting principal components satisfy regarding their relationship to one another in the feature space?

Accepted Answer

In Principal Component Analysis, the resulting principal components must be orthogonal to one another. Orthogonality means that each principal component is at a 90-degree angle to every other principal component in the feature space. This geometric property corresponds to a mathematical condition called zero correlation, meaning the components are uncorrelated and do not share any redundant information with one another. Because each component represents a different direction of variance in the data, being orthogonal ensures that the first component captures the largest possible spread of data, while each subsequent component captures the largest possible remaining spread that is completely independent of the previous ones. In practical terms, this allows you to condense a complex dataset into fewer dimensions without losing the unique patterns defined by the original variables. If you imagine a two-dimensional plot, the first principal component is a line drawn through the data along its longest stretch, and the second principal component is a line drawn perpendicular to the first, capturing the remaining vertical spread.

Home → All Courses → Science and Mathematics Courses → Data Science and Predictive Analytics → Flashcard

In Principal Component Analysis (PCA), what property must the resulting principal components satisfy regarding their relationship to one another in the feature space?