Dimensionality Reduction Techniques for Multivariate Categorical Data
Nikolaos Papafilippou, Zacharenia Kyrana, Emmanouil D. Pratsinakis, Christos A. Dordas, Georgios C. Menexes · 2025
High-dimensional categorical data challenges Data Science, Machine Learning, and Statistics in analyzing variability and interpretation. Using the “Forest Cover Type” dataset (n=581,012), this paper explores dimensionality reduction methods: PCA, MCA, CATPCA, FAMD, MFA, and NLCCA. Results show no method is universally best; the choice depends on data and goals. Advanced strategies like combining PCA, MCA, and CATPCA offer unique insights, emphasizing diverse approaches tailored to study objectives.