Product Size Recommendation for Fashion E-commerce
summary
Paper claims a probabilistic size-recommendation system that reduces size-related returns by inferring a latent customer true size and product true size and picking the closest variant. It compares a hinge-loss point-estimate, classifiers and a Bayesian model that uses variational inference to produce full posteriors. Evaluation uses tens of millions of transactions, shows better AUC for the Bayesian approach, and the algorithm is implemented in Python and a few GB memory with fast runtime.
AI-written summary · about
chapters (4)
- 0:00Project overview and goals
- 15:39Model evaluation and results
- 18:02Summary and future work
- 19:19Q&A session
related talks
35mSpotting trends and tailoring recommendations: PySpark on Big Data in fashion
24mAI-Powered Transformation in Fashion E-Commerce
37mFood is personal! Design and impact of personalized recommendations in e-grocery
17mCalculating recommendations based on product images #bbuzz
30mAI in Retail - Advancing Promo Demand Forecasting
43mRecommendation System beyond traditional Collaborative filtering
28mThe Power of Personalized Customer Experiences Through AI Innovations
18mSynthesizing human and machine capabilities