Evaluating research centers requires assessing both scientific output and the nature of collaborations among researchers. This study adopts a supervised machine learning approach to explore the collaborative dynamics at the Italian Institute of Technology, a leading research organization in Italy. Using data from 1,469 researchers at the Institute’s headquarters in 2020, I examine co-authorship networks formed through joint publications. This research identifies the relevance of network statistics (endogenous variables) and authors’ social and demographic attributes (exogenous variables) in shaping collaboration. By integrating network-based and individual-level characteristics, this study advances the understanding of link formation, a critical topic in knowledge and innovation networks.