摘 要
社交网络平台的快速发展使得用户生成内容呈指数级增长，如何精准地为用户提供个性化推荐成为亟待解决的问题。本研究旨在构建基于图神经网络的社交网络推荐系统，以提高推荐的准确性和多样性。传统推荐算法在处理社交网络数据时面临诸多挑战，如数据稀疏性、冷启动问题以及难以捕捉复杂的社交关系等。为此，提出一种融合多源信息的图神经网络模型，该模型不仅考虑了用户的直接交互行为，还引入了社交关系和内容特征作为补充信息，通过图结构有效传播节点间的潜在关联。实验采用真实社交网络数据集进行验证，结果表明所提方法在多个评价指标上显著优于现有主流推荐算法，特别是在长尾项目推荐方面表现出色，有效缓解了冷启动问题。此外，通过对模型内部机制的深入分析，发现其能够自动学习到不同类型的社交关系对推荐结果的影响权重，这一特性为理解社交网络中的信息传播规律提供了新的视角。本研究的主要贡献在于创新性地将图神经网络应用于社交网络推荐领域，提出了有效的解决方案，并为后续研究奠定了理论基础。
关键词：图神经网络；社交网络推荐系统；多源信息融合；冷启动问题；长尾项目推荐
Abstract
The rapid development of social network platforms has led to an exponential increase in user-generated content, making it imperative to address the challenge of providing accurate personalized recommendations to users. This study aims to construct a social network recommendation system based on graph neural networks (GNNs) to enhance both the accuracy and diversity of recommendations. Traditional recommendation algorithms face numerous challenges when processing social network data, including data sparsity, cold start problems, and difficulties in capturing complex social relationships. To this end, we propose a multi-source information integrated GNN model that not only considers direct user interactions but also incorporates social relationships and content features as supplementary information, effectively propagating latent associations between nodes through the graph structure. Experiments conducted using real-world social network datasets demonstrate that the proposed method significantly outperforms existing mainstream recommendation algorithms across multiple evaluation metrics, particularly in long-tail item recommendations, thereby effectively alleviating the cold start problem. Furthermore, an in-depth analysis of the model's internal mechanisms reveals its capability to automatically learn the influence weights of different types of social relationships on recommendation outcomes, offering new insights into the patterns of information diffusion within social networks. The primary contribution of this research lies in innovatively applying GNNs to the domain of social network recommendations, proposing effective solutions, and laying a theoretical foundation for future studies.
Keywords：Graph Neural Network；Social Network Recommendation System；Multi-source Information Fusion；Cold Start Problem；Long-tail Item Recommendation
目 录
摘 要 I
Abstract II
引 言 1
第一章 图神经网络基础与应用 2
1.1 图神经网络概述 2
1.2 社交网络中的图结构 2
1.3 图神经网络在推荐系统中的优势 3
第二章 社交网络数据建模 5
2.1 用户关系图构建 5
2.2 多模态数据融合 5
2.3 动态社交图更新机制 6
第三章 推荐算法设计与优化 8
3.1 基于社交网络推荐系统 8
3.2 用户兴趣传播模型 8
3.3 算法性能优化策略 9
第四章 系统实现与评估 11
4.1 推荐系统架构设计 11
4.2 实验环境与数据集 11
4.3 性能评估与结果分析 12
结 论 14
参考文献 15
致 谢 16
