CMStatistics 2018: Start Registration
View Submission - CMStatistics
B0972
Title: Clustering and visualizing large cattle-trading networks using self-organizing maps Authors:  Madalina Olteanu - Pantheon-Sorbonne University (France) [presenting]
Kevin Pame - MaIAGE INRA Universite Paris Saclay (France)
Gael Beaunee - Bioepar INRA (France)
Caroline Bidot - MaIAGE INRA Universite Paris Saclay (France)
Elisabeta Vergu - MaIAGE INRA Universite Paris Saclay (France)
Abstract: Networks have drawn quite a burst of attention in the last years, and two of the related questionings are understanding the underlying structure(s) of the network and visualizing simplified version(s) of it. When one aims at bringing into light how the groups of entities in a graph are organized and how they interact, clustering and visualization were proven to be very useful. More particularly, the use of a recent relational version of the self-organizing maps (SOM) algorithm provides a unified tool for both purposes, while allowing for a wide range of alternatives in terms of assessing the similarity between vertices and/or edges. The aim is to adapt a bagged version of relational SOM for time-varying networks, and to explore the French cattle-trading network. The network is represented as a dynamical graph with a daily resolution level, where the vertices are the farms (and the commercial operators), and the edges are represented by the animal exchanges. The graph is directed (from sellers to buyers), weighted (by the number of exchanged animals), and time varying (a transaction occurs at a given time-instant). Furthermore, additional information on the vertices, such as the geographical situation, the type of farm, and so on, is used either during the clustering procedure or for validating the results.