Diversity Preference-Aware Link Recommendation for Online Social Networks.

This article focuses on the development of a diversity preference-aware link recommendation (DPA-LR) method for online social networks, addressing the limitation of existing link recommendation approaches that overlook individual users' varying preferences for friend diversity. The study defines and operationalizes "diversity preference" as a user's inclination toward homophily (similarity) or heterophily (diversity) across multiple profile dimensions (e.g., major, school, employer), and formulates the link recommendation problem as an optimization task to select candidate friends that best match each user's diversity preference at the dimension level. The proposed method employs an iterative optimization algorithm to solve this NP-hard problem and is empirically evaluated on large-scale Google+ data, demonstrating significant improvements over state-of-the-art link recommendation and diversification benchmark methods in both satisfying users' diversity preferences and recommendation accuracy (precision, recall, F1 score). The study highlights implications for social network operators and recommender system design, suggesting that personalized consideration of diversity preference can enhance user experience and network connectivity.