22/10/2019 – 22/10/2019 – Graph-Based Data Mining for the Economy

Ateneo Università Genova: 22/10/2019 – 22/10/2019 – Graph-Based Data Mining for the Economy

Abstract:
The wealth of digitized, structured information which has been made available to researchers has started the “big data” era – this is also true for socio-economic disciplines. Economy has not missed the opportunity, and is benefitting from novel approaches which go beyond traditional econometrics.
Complex networks lend themselves very naturally to capturing and expressively representing diverse economic phenomena. A prime example, based on Japanese data from the Toyo Keizai series, will be subject to a graph-theoretical treatment, and valuable insights will result in a discussion on Japanese keiretsu-like shareholdings. In doing this, a simple community detection algorithm will be presented.
The elusive concept of communities will allow to introduce a novel, very specific kind of network metric, namely the so-called Bridgeness Centraliy, also offering powerful insights as to the relevance of certain economic actors embedded in a networked environment, specifically, the world air transport system.
Finally, a bleeding-edge graph-based, semi-supervised machine learning application will be offered as a successful interdisciplinary effort in extracting information from the interconnected nature of economic systems, generalizing and extending the standard PageRank algorithm.
Keywords: Network Science, Machine Learning, Data Mining
Sarah de Nigris | Universität Koblenz-Landau
Dr. Sarah de Nigris research revolves on dynamical processes upon networks and graph-based machine learning. After a PhD in Theoretical and Mathematical Physics at the Centre de Physique Théorique in Marseilles, France, she worked as a postdoctoral researcher at the Mathematics Department in Namur (BE) and at the LIP institute at the École Normale Supérieure de Lyon (FR). She is currently a postdoctoral researcher at the WeST institute in Koblenz (DE), in the frame of the E-Democracy project, working on machine learning techniques to be applied on on large social data.
Matteo Morini | Università di Torino, École Normale Supérieure de Lyon, Universität Koblenz-Landau
Matteo Morini got his PhD at the École Normale Supérieure de Lyon, France; he works on complex networks, and develops agent-based and econometric models at the University of Torino. He is also employed at the Institut für Wirtschafts- und Verwaltungsinformatik at the Universität Koblenz-Landau, teaches courses on complex systems in the Carlo Alberto postgraduate programme and sits in the board of directors (as vice-president) of the Swarm Development Group. He co-authored and co-edited two books on complexity and ABMs. For more of his work, see http://perso.ens-lyon.fr/matteo.morini
Fonte: Ateneo Università Genova