projects/factoclass

FactoClass: letting data reveal their own structure

FactoClass was part of my early work as a Statistics student at the Universidad Nacional de Colombia, developed under the guidance of Campo Elías Pardo and inspired by Lebart et al.’s methodology. It reflects how deeply exploring data structure enables meaningful clusters.

Statistics DataScience Multivariate Analysis Clustering RStats Applied Statistics
FactoClass overview (EN)

Exploration first

I’ve always believed that before classifying, we need to explore. FactoClass comes from that idea.

This work is part of what I did while I was a Statistics student at the Universidad Nacional de Colombia. It was my first steps as a statistician and programmer, developed under the guidance of Campo Elías Pardo and inspired by Lebart and colleagues.

Method: factorial + clustering

The package implements a strategy that combines factorial methods and cluster analysis to explore complex data tables. First, a factorial space organizes variability and reduces noise; then clustering follows—initial partitions, Ward hierarchical aggregation, and a final K-means consolidation—all within that shared space.

Applications

FactoClass is useful when the objective is exploratory rather than predictive, and when understanding internal structure comes before modeling. It fits complex multivariate tables (quantitative, categorical, or frequency) across social, demographic, territorial, or survey-based datasets. It helps identify meaningful groups while preserving interpretability, making it applicable to population profiling, market segmentation, behavioral analysis, academic research and teaching focused on multivariate exploration, and more.

Why it matters

Looking back, this work reminds me that deeply understanding and exploring the nature of things is what allows us to classify them in a meaningful way.