Researchers have developed a new framework called Hierarchical GTV Estimation (HGE) to tackle the challenge of estimating Gross Transaction Value (GTV) in e-commerce platforms that use a container-based display paradigm. This framework addresses issues like high label variance, label sparsity, and dynamic item heterogeneity by employing a Hybrid Set-Aware Encoder (HSAE) for modeling item composition and feature interactions. HGE also incorporates a Proxy-Label Learning (PLL) strategy to decompose GTV and a Residual Correction and Ensemble Fusion (RCEF) module to combine item-level and container-level inference. The system has been deployed and demonstrated significant improvements in online A/B tests, increasing Revenue Per Search (RPS) by 1.42%. AI
IMPACT This research offers a novel approach to improving transaction value estimation in e-commerce, potentially impacting how online platforms optimize sales and user experience.
RANK_REASON Academic paper detailing a new framework and its experimental results. [lever_c_demoted from research: ic=1 ai=0.7]
Read on arXiv cs.IR (Information Retrieval) →
- Gross Transaction Value
- Hierarchical GTV Estimation
- Hugging Face
- Hybrid Set-Aware Encoder
- Proxy-Label Learning
- Residual Correction and Ensemble Fusion
- Revenue Per Search
- XAUC
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →