PCA with Missing Values and Relaxed Solution of k-Means Clustering

Abstract

It has shown that a relaxed solution of k-Means clustering can be derived by a PCA-guided procedure. This paper considers k-Means-like clustering of incomplete data based on PCA handling missing values. Besides solving the eigenvalue problem of covariance matrices, k-Means-like partitions are derived through lower rank approximation of data matrix ignoring missing elements.