|
Abstract
|
In the current data-driven era, recommendation systems, as a vital component of online platforms, play a key role in shaping users’ experiences and guiding their decision-making in areas such as e-commerce, social networks, and entertainment systems. However, with the increasing scale and complexity of interactions in modern digital ecosystems, traditional recommendation algorithms face several challenges such as severe data fragmentation, cold start problem, and lack of explainability. Graph-based methods have been proposed as a powerful approach to modeling the relationships between users and items in complex network structures, but their reliance on labeled data and high computational cost limit their generalizability and scalability. Meanwhile, self-supervised graph learning and especially graph contrastive learning have provided a new field for extracting meaningful representations by eliminating the dependence on labels, but most of these methods neglect to consider the community-based structures that fundamentally shape users' behavior and preferences. Integrating social information and community structures in the contrastive learning process can reveal hidden and semantic relationships in the network and increase the stability of representations in sparse data conditions; as a result, it can improve the accuracy and explainability of recommendation systems. In addition, with the increasing growth of data volume and the need for real-time processing in large-scale recommendation systems, the need to develop efficient and community-based frameworks that can reduce computational costs while maintaining high accuracy is increasingly felt. Therefore, this research aims to fill the gap at the intersection of community detection, self-supervised adversarial learning, and efficient graph representation for recommendation systems in complex networks; an approach that, from a theoretical and practical perspective, can be considered an effective step towards developing
|