Graphical models for complex networks: an application to Italian museums

This paper applies probabilistic graphical models in a new framework to study association rules driven by consumer choices in a network of Italian museums. The network consists of the museums participating in the programme of Abbonamento Musei Torino Piemonte, which is a yearly subscription managed by Associazione Torino Città Capitale Europea. It is available to people living in the Piemonte region, Italy. Consumers are card-holders, who are allowed entry to all the museums in the network for one year. We employ graphical models to highlight associations between the museums driven by card-holder visiting behaviour. We use both simple undirected graphs and more complex directed graphs, and we do not make any hypothesis on the models but rather learn their structures directly from the data. We also use methodologies and tools for robust network identification and principal component analysis to complete the analysis of the phenomenon.