Skip to Main content Skip to Navigation
Master Thesis

Analyse des correspondances multiples parcimonieuses

Abstract : This report presents the Sparse Multiple Correspondence Analysis (SMCA), the sparse equivalent of Multiple Correspondence Analysis (MCA). MCA is a widely used method for exploratory analysis of categorical data. Adding sparsity to an MCA improves its interpretability, especially for the analysis of large data. Based on singular value decomposition (SVD), MCA can be made sparse by generalizing the constrained SVD algorithm (CSVD). More specifically, CSVD requires two additional properties: to take into account the mass and weight matrices that are characteristic of the MCA (weight of individuals and frequency of modalities) and to select entire groups of variables (one group consisting of the complete disjunctive coding of a categorical variable) or of observations. We illustrate the SMCA with two sets of data: the first is a questionnaire on perception and knowledge of Maroilles and the second is a set of genetic data taken from a study on Alzheimer's disease. These two examples allow us to show the interest of the sparsity constraint for the interpretation of the estimated dimensions, as well as for the exploration of the links between individuals and modalities.
Document type :
Master Thesis
Complete list of metadata
Contributor : Agrocampus Ouest Connect in order to contact the contributor
Submitted on : Thursday, October 15, 2020 - 2:41:48 PM
Last modification on : Sunday, January 10, 2021 - 3:08:50 AM


Files produced by the author(s)


  • HAL Id : dumas-02968178, version 1



Julie Le Borgne. Analyse des correspondances multiples parcimonieuses. Sciences du Vivant [q-bio]. 2020. ⟨dumas-02968178⟩



Record views


Files downloads