Reproduce “Large-scale Real-time Personalized Similar Product Recommendations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhi Liu, Yan Huang, Jing Gao, Li Chen, Dong Li
Researchers developed a large-scale real-time personalized algorithm for similar product recommendations that integrates dynamic user interests and explicitly optimizes for add-to-cart conversions. Deployed on a major e-commerce platform, the enhanced model achieved approximately a 10% increase in online add-to-cart numbers while processing 2,000 requests per second.
Key insights
- Integrating real-time user interest modeling, based on a dynamic user vector from recent interactions, significantly improves the relevance of similar product recommendations.
- Explicitly weighting add-to-cart events in the loss function during model training is a powerful strategy to optimize recommender systems directly for conversion metrics like add-to-cart rate.
- A hybrid architecture combining offline candidate generation with online real-time ranking is effective for achieving both scalability and personalization in large-scale e-commerce recommendation systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.