The movie recommendation algorithm based on the TransD model and AIGC empowerment and its application effectiveness...

Area of Science:

• Artificial Intelligence
• Information Retrieval
• Data Science

Background:

• Traditional recommendation systems struggle with cold start, limited semantic understanding, and user interest representation.
• Existing models often rely on manual tags and fail to fully exploit structured information or diverse user interests.

Purpose of the Study:

• To enhance recommendation systems by addressing cold start issues, improving semantic understanding, and modeling user interest diversity.
• To propose a novel framework integrating Knowledge Graph Embedding (TransD) and Artificial Intelligence Generated Content (AIGC) for movie recommendations.

Main Methods:

• Utilized TransD for dynamic semantic modeling of heterogeneous entities and relationships within knowledge graphs.
• Employed AIGC to extract latent interest dimensions, emotional characteristics, and semantic tags from user reviews for profile construction.
• Constructed a content tag completion system and a high-dimensional user interest profile.

Main Results:

• The proposed model achieved superior performance across MovieLens datasets, with hit rates up to 0.878 and Mean Average Precision (MAP) up to 0.637.
• Demonstrated high user satisfaction scores (up to 0.89) and click-through rates (CTR) (up to 0.35), significantly outperforming traditional models.
• Showcased superior stability and semantic adaptability, particularly for cold start users and interest transitions.

Conclusions:

• The integrated approach effectively combines structured and unstructured information for advanced movie recommendations.
• The study offers significant theoretical and practical contributions to intelligent recommendation systems, knowledge graph embedding, and AIGC-based hybrid modeling.