On the Effect of Data Dimensionality on Eigenvector Centrality

TL;DR

This paper investigates how the dimensionality reduction of multi-relational data into traditional graphs affects eigenvector centrality, revealing that centrality rankings can change depending on the hypergraph's order.

Contribution

It introduces the first known hypergraph example where eigenvector centrality varies with the matrix order, highlighting the impact of data dimensionality on spectral analysis.

Findings

Eigenvector centrality can differ based on the hypermatrix order.

Dimensionality reduction may lead to different centrality conclusions.

Hypergraph analysis reveals properties not captured by traditional graphs.

Abstract

Graphs (i.e., networks) have become an integral tool for the representation and analysis of relational data. Advances in data gathering have lead to multi-relational data sets which exhibit greater depth and scope. In certain cases, this data can be modeled using a hypergraph. However, in practice analysts typically reduce the dimensionality of the data (whether consciously or otherwise) to accommodate a traditional graph model. In recent years spectral hypergraph theory has emerged to study the eigenpairs of the adjacency hypermatrix of a uniform hypergraph. We show how analyzing multi-relational data, via a hypermatrix associated to the aforementioned hypergraph, can lead to conclusions different from those when the data is projected down to its co-occurrence matrix. In particular, we provide an example of a uniform hypergraph where the most central vertex (\`a la eigencentrality) changes depending on the order of the associated matrix. To the best of our knowledge this is the first known hypergraph to exhibit this property.