Knowledge Graph-based Recommender Systems: A Case Study on Movie Domain

The proliferation of digital content has heightened the need for effective recommendation systems, particularly in the movie industry. Traditional recommendation methods like collaborative and content-based filtering often struggle with data sparsity and cold start problems. To address these challenges, we propose a Knowledge Graph (KG)-based movie recommendation system that leverages the rich relational information inherent in KGs. By employing the Translating Embeddings (TransE) model, we embed entities (movies, actors, genres, users) and relationships into a continuous vector space, capturing complex interactions and semantic meanings. We validate our approach using the MovieLens-lM dataset, constructing a KG that integrates movie metadata and user interaction data. Experimental results demonstrate that our KG-based system significantly outperforms traditional baseline models regarding precision, recall, and F1-score. Ablation studies further reveal the critical contributions of different KG components to the overall performance. This research showcases the potential of KGs in enhancing recommendation quality and provides a robust framework for future applications in various domains.