Applying separative non-negative matrix factorization to extra-financial data

TL;DR

This paper introduces an innovative application of non-negative matrix factorization (NMF) to analyze highly correlated extra-financial data, demonstrating improved clustering over PCA and enhanced results with a preliminary data separation step.

Contribution

The paper presents a novel use of NMF for extra-financial data, including a data separation step that enhances clustering quality compared to traditional PCA.

Findings

NMF provides better clustering than PCA for correlated data.

Preliminary data separation improves clustering results.

Application to extra-financial data shows practical effectiveness.

Abstract

We present here an original application of the non-negative matrix factorization (NMF) method, for the case of extra-financial data. These data are subject to high correlations between co-variables, as well as between observations. NMF provides a much more relevant clustering of co-variables and observations than a simple principal component analysis (PCA). In addition, we show that an initial data separation step before applying NMF further improves the quality of the clustering.