Towards personalized recommendation with enhancing preference matching through scene-weighted reranking

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Recommendation systems are vital for user engagement.
• Existing reranking algorithms often overlook local item relationships within subsets.
• Pairwise item interactions are the primary focus of current methods.

Purpose of the Study:

• To introduce a novel reranking algorithm that captures local item relationships.
• To address the limitations of existing methods in understanding complex item interactions.
• To enhance the accuracy and quality of personalized recommendations.

Main Methods:

• Introduced the concept of "scenes" to mine local relationships among multiple items.
• Represented inter-scene correlations using undirected graphs.
• Proposed a scene-weighted reranking algorithm integrating scene-user preference matching and item-scene similarities.

Main Results:

• The scene-weighted reranking algorithm achieved more accurate item rankings.
• The proposed method better reflects users' true preferences compared to existing approaches.
• Experimental results demonstrated higher-quality recommendation sequences.

Conclusions:

• The novel approach effectively captures both local and global item relationships.
• The scene-weighted algorithm enhances preference matching in personalized recommendation systems.
• This research offers a more nuanced understanding of item interactions for improved recommendations.