Overview

This project proposes a movie recommendation system that integrates knowledge graph information into matrix factorization to improve recommendation accuracy and address data sparsity issues.

By incorporating semantic relationships between entities (such as movies, genres, and actors), the system enhances traditional collaborative filtering and provides more context-aware recommendations.

Project Period

2022.09 – 2023.06

Development Tools

Python · Machine Learning · Matrix Factorization · Knowledge Graph · Neo4j

Role

End-to-end Development (Model Design + Implementation + Evaluation)

My Contributions

• Designed and implemented a hybrid recommendation model combining matrix factorization with knowledge graph features
• Integrated entity relationships (e.g., movies, genres, actors) into embedding space for improved recommendation quality
• Addressed data sparsity issues by leveraging external knowledge graph information
• Conducted model training and evaluation to compare performance with traditional collaborative filtering methods
• Analyzed recommendation results and optimized model parameters

Core features

• Matrix Factorization
  Learns latent user–item representations based on interaction data
• Knowledge Graph Integration
  Embeds semantic relationships between entities to enrich feature representation
• Hybrid Recommendation Strategy
  Combines collaborative filtering with knowledge-driven features for improved accuracy

project preview