Recommender Systems for Personalized Financial Advice
summary
Production recommender uses matrix factorization collaborative filtering trained on implicit holdings turned into confidence-weighted labels where trading activity becomes the best proxy for satisfaction; assets are grouped into product bins (≈13 asset features) to avoid cold start and track time‑variant characteristics. Explanations are generated as “As you bought A, you might like B” by finding block structures in the user–item matrix. Binary liked/disliked feedback sets extreme confidence and forces the loss to push categories up or down. System passes RM approval and runs in Bank J. Safra Sarasin’s e‑bank.
AI-written summary · about
chapters (7)
- 0:00Start
- 2:32Introduction
- 6:43Matrix Factorization Collaborative Filtering
- 14:24Asset Categories
- 15:37Time Variant Product Features
- 18:50The Feedback Loop
- 24:16Recap Why Do We Need a Recommender for Financial Advice
related talks
41mRecommender Systems: Crafting Effective Personalised Experiences
58mRecommender Systems – A Walkthrough
37mA primer on recommendation systems.
43mRecommendation Systems in Market Research
29mA content-based recommender system for financial news
28mThe Power of Personalized Customer Experiences Through AI Innovations
29mThe Challenges of Building a Personalized Recommendation System in E-Commerce
32mLet's talk about Recommendation Systems!