<p>The rapid growth of digital entertainment has highlighted the pivotal role of recommendation systems in helping users discover animation videos tailored to their preferences. However, the complex and evolving nature of user interactions on online platforms often leads to sparse datasets, limiting the effectiveness of traditional Collaborative Filtering (CF) methods. This issue is especially prevalent in animation recommendations, where diverse genres and varying user tastes intensify the cold-start problem, making it difficult to accurately predict preferences for new or inactive users. To address these challenges, we propose a novel Social Relationship-enhanced Collaborative Filtering (SRCF) algorithm that integrates user social relationships into the traditional CF framework. This approach generates additional deepfake user-item interaction data, enriching the model with more reliable social context. Additionally, we introduce AnimationTrustCF (ATCF), an advanced hybrid recommendation model that combines user-item interactions with social relationship metrics to improve the recommendation process. By leveraging multidimensional data, ATCF enhances the system’s resilience to data sparsity while improving both the accuracy and robustness of animation video suggestions. Through extensive experiments on a real-world dataset, we demonstrate that ATCF outperforms traditional models in overcoming data sparsity and offers a significantly improved user experience in discovering animation content.</p>