Principal component analysis is a method of summarizing the information in multidimensional data observed for features that are correlated with each other into new features expressed as a linear combination of the original features, with as little loss of information as possible.

The data to be classified by machine learning is often highly dimensional, well beyond three dimensions. This makes data visualization difficult and computationally expensive.

Even in such cases, principal component analysis can be used to compress dimensions and project the data into a 1D line, 2D plane, or 3D space to visually grasp the data structure.

This article describes the basic theory behind principal component analysis.

👇
https://laid-back-scientist.com/en/pca-theory