Reproduce “Deep Evolutional Instant Interest Network for CTR Prediction in Trigger-Induced Recommendation” 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
Zhibo Xiao, Luwei Yang, Tao Zhang, Wen Jiang, Wei Ning, Yujiu Yang
A novel deep learning model, Deep Evolutional Instant Interest Network (DEI2N), precisely models the dynamic evolution of a user's instant interest and integrates temporal information for click-through rate prediction in trigger-induced recommendation systems. This approach achieved superior performance across datasets, including a 0.56% to 1.53% conversion rate improvement in online A/B testing on Alibaba.com.
Key insights
- User interest in Trigger-Induced Recommendation (TIR) is not static but dynamically evolves, particularly as users scroll through recommendation lists.
- Explicitly modeling the "instant interest" ignited by a trigger item, and its dynamic change, is crucial for accurate CTR prediction in TIR.
- Temporal information in user behavior sequences (order and intervals) provides valuable context for understanding evolving preferences and improving attention mechanisms.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.