About Singular Value Decomposition

SVD and the closely-related PCA are well established feature extraction methods that have a wide range of applications. Oracle Machine Learning for SQL implements Singular Value Decomposition (SVD) as a feature extraction algorithm and Principal Component Analysis (PCA) as a special scoring method for SVD models.

SVD and PCA are orthogonal linear transformations that are optimal at capturing the underlying variance of the data. This property is very useful for reducing the dimensionality of high-dimensional data and for supporting meaningful data visualization.

SVD and PCA have a number of important applications in addition to dimensionality reduction. These include matrix inversion, data compression, and the imputation of unknown data values.