Online social networks (OSNs), such as Twitter and Facebook, enable users to create, share, and interact with diverse content, thereby producing intricate pathways for information propagation. This flow, which can be modeled through graphs that capture Follower/Following relationships and various interactions such as retweets and mentions, can offer valuable insights into the dynamics of online social behavior and information sharing. While the Follower/Following networks are important for modeling user characteristics and behaviors, their construction can prove expensive in terms of both time and resources. More importantly, in some OSNs, partial or full restrictions have been posed on the access to users’ Follower/Following information, effectively rendering the regular construction process of Following graphs intractable. In this paper, we explore the viability of extracting users’ Following connections from their Retweet/Mention networks through predictive models. Taking Twitter as a case study, we train and contrast the performance of five different models, including classical Machine Learning (ML) methods as well as a recently developed Deep Learning (DL) approach, on two different datasets. The difference in prediction results across the models and datasets is traced and analyzed. Lastly, we round up the contributions by providing a carefully curated Twitter dataset compiled from over 9,000 individuals’ timelines, encapsulating their retweets, followers, and following networks. Taken together, the results and findings featured herein can aid in paving the way for improved understanding and modeling of online social networks.