ECCO 2024
Abstract Submission

74. New methodology for Hypergraph Clustering

Invited abstract in session FC-1: Graphs, stream Graphs.

Friday, 13:30 - 15:00
Room: L226

Authors (first author is the speaker)

1. Francisco Temprano Garcia
Stats and OR, Universidad de Sevilla
2. Stefano Benati
School of International Studies, University of Trento
3. Justo Puerto
Estadistica e I.O., Universidad de Sevilla

Abstract

This talk deals with the problem of partining set of nodes into highly connected subsets over hypergraphs. Due to the lack of universal consensus and developed theory in literature, we must come up with a formal definition aim of the problem, in order to propose new optimization models to solve it. A hypergraph is known to be a generalization of a graph that allows us to represent a huge amount of real-life interactions between elements of a data sample that the classic and original networks can not. Once we formally define the problem, we develop a large list of functions able to measure the goodness of the node set subdivision. Next, we compare all these methods by using of mathematical programming and extensive computational experiments. Thus, we can conclude which methods describe better the properties of partition structures and which ones perform better, since multiple compact formulations and column generation algorithms are developed to solve the partitioning hypergraph problem. Finally, we have applied all methods to Eurobarometer data, showing the applicability of the introduced methodology.

Keywords

Status: accepted

Back to the list of papers