Even if for most people playing video games is a healthy leisure activity, a minority of vulnerable users present an excessive use associated to negative consequences (e.g., psychosocial maladjustment, sleep interference) and functional impairment. The current study first aims to identify psychological factors that contribute to discriminate highly involved (but healthy) gamers from problematic gamers. For that purpose, we used a cluster analysis approach to identify different groups of gamers based on their profiles of passion towards gaming (using the Dualistic Model of Passion). Another objective of the present study is to explore, using supervised machine-learning, how gaming disorder symptoms, assessed within the substance use disorder framework (e.g., tolerance, withdrawal), might be linked to harmonious and/or an obsessive passion for gaming. Three distinct clusters of gamers were identified based on their passion profiles, including risky gamers, engaged gamers, and casual gamers. Supervised machine-learning algorithms identified that specific gaming disorder symptoms (salience, mood modification, tolerance, low level of conflict) were predominantly related to harmonious passion, whereas others (withdrawal, high level of conflict, relapse) were more directly related to obsessive passion. Our results support the relevance of person-centered approaches to the treatment of problematic gaming.