Building your first recommender
summary
Builds a recommender from user events: capture clicks/plays, enrich and store them, then train collaborative filters using Jaccard/Manhattan/L2 and cosine similarity with rating normalization to reduce bias. Uses LensKit as the Java library, a Spring backend and React frontend, an in‑memory event collection, and a demo dataset of the rolling 500 albums. Adds k‑means clustering to scale and a content‑based fallback for cold users; the demo produces live recommendations and a dashboard showing coverage.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 2:10Advertorial vs recommendation
- 6:07Recommendation goals and types
- 7:52Collaborative filtering basics
- 12:49Distance metrics explained
- 16:51Cosine similarity and normalization
- 19:18Finding user neighborhoods
- 22:26Demo time and predictions
- 25:59Building the recommender app
- 28:05LensKit library overview
- 34:23Scaling with machine learning
- 36:49Content-based recommendations
- 38:15AI in recommendations
- 40:04Q&A and next steps
related talks
41mRecommender Systems: Crafting Effective Personalised Experiences
33mPicking What to Watch Next: Build a Recommendation System
58mRecommender Systems – A Walkthrough
55mBuilding a Recommendation Engine with Machine Learning Techniques
33mPeople who liked this talk also liked ... Building Recommendation...
01h 31mHands on - Build a Recommender system
31mCreating recommender systems with Spark, Scala and Prediction.Io
20mBuilding Recommender system