Projection Sparse Principal Component Analysis: an efficient method for improving the interpretation of principal components Giovanni Maria Merola a,∗ a Xi’an Jiaotong-Liverpool University Abstract Sparse principal components analysis (SPCA) methods approximate princi- pal components with combinations of few of the observed variables. Sparse components are more interpretable than standard principal components as they identify few key features of a dataset. We propose a practical SPCA method in which sparse components are computed by projecting the full principal components onto a subset of the variables. We show that these components explain more than a predetermined percentage of the variance explained by the principal components. We also show that this approach is strictly related to least squares SPCA by providing a novel interpretation for the latter. We propose a simple and efficient algorithm that uses sim- ple forward selection to select variables and the power method to compute eigenvectors. We illustrate the method with the analysis of a real dataset con- * Department of Mathematical Sciences. 111 Renai Road, Dushu Lake Higher Edu- cation Town, Suzhou Industrial Park, Suzhou, Jiangsu Province, PRC 215123. Email address: giovanni.merola@gmail.com (Giovanni Maria Merola) Preprint submitted to Elsevier December 3, 2016