A primer on recommendation systems.
summary
Compares concrete approaches: content filtering (feature vectors from items) versus collaborative filtering (user-behavior features), shows that matrix factorization reduces sparsity and gives examples that actually work. Implements demos with MovieLens ratings and a Reddit dataset of 1.8M comments, using pandas/numpy/scipy and SQLite, builds an item-item Reddit recommender and a user-item MovieLens recommender, runs truncated SVD k=100 to reconstruct ratings, and reports plausible outputs (Transformers/Star Wars for an action fan, Disney-heavy picks for an animation fan).
AI-written summary · about
chapters (13)
- 0:00Introduction
- 3:20What are recommendation systems
- 4:22Similarity in recommendation systems
- 6:08Recommendation system approaches
- 6:49Content filtering examples
- 9:38Content-based movie recommender
- 13:32Collaborative filtering approaches
- 17:04Item-to-item collaborative filtering Reddit
- 21:38User-item collaborative filtering MovieLens
- 27:21Hybrid models
- 28:25Python libraries
- 29:07Singular Value Decomposition
- 35:15Q&A
related talks
32mLet's talk about Recommendation Systems!
33mPicking What to Watch Next: Build a Recommendation System
20mBuilding Recommender system
01h 38mBuilding Recommender Systems Using Python
58mRecommender Systems – A Walkthrough
01h 31mHands on - Build a Recommender system
01h 21mUnlocking Personalization: A Deep Dive into Modern Recommendation Algorithms
19mEvaluating narrative-driven movie recommendations on Reddit