GRAPH NEURAL NETWORKS FOR SOCIAL NETWORK ANALYSIS: DETECTING FAKE PROFILES & BOTNETS

Abstract

Social networks have become one of the most significant sources of interconnected data in the digital era, enabling communication, information sharing, and collaboration among millions of users worldwide. The analysis of social networks involves understanding complex relationships, community structures, user interactions, influence propagation, and behavioral patterns. Traditional machine learning techniques often struggle to effectively model the highly interconnected and non-Euclidean nature of social network data. To address these challenges, Graph Neural Networks (GNNs) have emerged as a powerful deep learning framework capable of learning from graph-structured data by capturing both node features and topological relationships. This project presents a Graph Neural Networksbased Social Network Analysis system designed to extract meaningful insights from social network graphs. In the proposed approach, users are represented as nodes, while their interactions, friendships, or connections are represented as edges within a graph structure. The system employs advanced GNN architectures such as Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), and GraphSAGE to learn hidden patterns and generate informative node embeddings. These embeddings are utilized for various analytical tasks, including community detection, user classification, link prediction, influence analysis, and recommendation systems.