Build a Recommender System in Apache Spark and Integrate it Using Akka
summary
Builds a recommender with Apache Spark using ALS matrix factorization, trains a rank-10 model with alternating least squares (about 20 alternations; heavy inner optimization he describes as ~100,000 steps per rating), exposes the model via an Akka HTTP microservice with a background training actor and an optional in-memory model, and logs recommended lists plus click events to compute precision/recall. Claims surprising real-world performance (he reports roughly 80% accuracy and an example of 70% precision / 66% recall), recommends regular full retrains and monitoring.
AI-written summary · about
chapters (7)
- 0:00Start
- 13:43Apache Spark Basics
- 19:31Building the recommender system
- 21:58Integrating everything into a micro service
- 26:58Three easy non-scientific steps
- 27:58The number crunching
- 33:49Try it yourself!
related talks
31mCreating recommender systems with Spark, Scala and Prediction.Io
40mIntroduction to Machine Learning
40mBuilding your own Distributed Streaming Recommendation Engine
26mLarge-Scale Recommendation System with Python and Spark
29mThe Challenges of Building a Personalized Recommendation System in E-Commerce
01h 48mRecommendation engines explained
15mWhat they didn’t teach you in Grad School #bbuzz
32mLeveraging recommender systems to personalise search results