Detection of Communities of Agents Interacting through Regional Innovation Policies

The detection of communities of agents that interacted over time through regional innovation policies is analyzed through the application of three methodologies: Clique Percolation Method (CPM) by Palla et al. (2005), Infomap by Rosvall and Bergstrom (2008), and Dynamic Cluster Index analysis (DCI) by Villani et al. (2013). The case study regards the policy interventions implemented by region Tuscany (Italy) in 2000-2006 with the aim of supporting innovative network projects among local actors. In a context of analysis centered on such a complex object as innovation, and affected by discontinuous temporal dynamics and changing configurations of partnerships of agents, the three methodologies are applied to investigate different specific aspects of community organizations aimed at developing innovative activities. For every methodology three models are developed. In CPM, the elaboration of three models following the observation of the features of all possible partitions makes it possible to overcome the problematic definition of the value of k. In Infomap, the observation of the chronological order in which funded projects were carried out is used to impose different restrictions on the circulation of simulated flows. Finally, the application of DCI analysis to a socio-economic context is developed through the elaboration of different variables describing agents’ behavioral profiles, and through an original contribution in using a cluster analysis aimed at coping with the large quantity of results that the algorithm produces. The investigation of relational structures (through CPM), of shared processes (through Infomap) and of integrated behaviors (through DCI analysis) allowed the identification of communities that reveal, respectively, meaningful characterizations in terms of agents’ participations in specific waves of the policy, of agents’ participations in projects operating in particular technological domains, and in terms of agents’ institutional typologies.