ETD

• academic ecosystem analysis
• bibliometric data
• explainable machine learning
• institutional research networks
• internal migration
• research flow prediction
• scientific mobility

The present research investigates the institutional positioning within the Italian national research ecosystem through the development and validation of BiFlow (Bibliometric Framework for Monitoring Institutional Research Flows). This framework introduces an analytical approach to monitor and predict the movement of academics out and in of research institutions, by leveraging bibliometric datasets.
The methodology adopts an observational and quantitative lens, integrating diverse data dimensions like scholarly output, institutional size through time, geographical distribution, and status-based indicators, to characterize institutional trajectories. The proposed workflow is structured as an end-to-end pipeline with four main steps. Starting with data engineering, this phase is articulated into data acquisition, cleaning, and feature extraction. It then moves to a predictive modelling phase aimed at classifying institutional flows (inflow, outflow, stability), followed by the integration of explainability post-hoc techniques to interpret model decisions and uncover the main drivers of research mobility. Bi-Flow also leverages networks analysis to understand the geographical direction of the flows in the Italian territory, and assess patterns of centrality and marginalization on the regional level.