Mutual Friend Recommendation in MSNs Exploiting Multi-Source Information Using a Two-Stage Deep Learning Framework

Abstract: Friendship inference in social networks has become a significant research area because of the proliferation of social media platforms and the valuable insights they offer. This study proposes a novel approach to infer friendships by exploiting multi-source information using a two-stage deep learning framework. The first stage several data sources, such as user interactions, profile information, and shared content, to generate comprehensive feature representations. The second stage employs a deep learning model to analyze these representations and predict friendship links with high accuracy. Results from experiments show that our approach outperforms traditional approaches, offering improved precision and recall in friendship inference. This research offers a strong basics for enhancing social network analysis and may leveraged for various applications like recommendation systems, targeted advertising, and community detection.