MNI: An enhanced multi-task neighborhood interaction model for recommendation on knowledge graph.

Area of Science:

• Artificial Intelligence
• Data Science
• Computer Science

Background:

• Recommendation systems often suffer from data sparsity and cold start problems.
• Knowledge graphs offer a way to incorporate side information for improved recommendations.
• Existing methods may not fully capture complex relationships within knowledge graphs.

Purpose of the Study:

• To propose an enhanced multi-task neighborhood interaction (MNI) model for knowledge graph-based recommendations.
• To address data sparsity and cold start issues in recommendation systems.
• To improve recommendation performance by leveraging richer structural information.

Main Methods:

• The MNI model integrates user-item interactions with neighbor-neighbor interactions.
• It utilizes semantic embeddings for entities and relations within the knowledge graph.
• A cross&compress unit enables shared latent features between items and knowledge graph entities for high-order interaction analysis.

Main Results:

• MNI effectively captures sophisticated local structures in knowledge graphs.
• The model demonstrates superior performance compared to state-of-the-art baselines.
• Experiments show significant improvements in both click-through rate (CTR) prediction and top-N recommendation tasks.

Conclusions:

• The proposed MNI model offers a robust solution for knowledge graph-enhanced recommendations.
• It successfully mitigates data sparsity and cold start challenges.
• MNI provides a promising direction for future research in recommendation systems.