Recommender Systems in Practice
From Simple Rules to Learned Rankings
- Pages
- 81
- Chapters
- 54
- Level
- Intermediate
- Format
- E-book
About this book
Recommendations drive what people watch, read and buy. This book explains how recommender systems work and how to build one that serves both users and the business.
You will start with simple baselines such as popularity and co-occurrence, then build collaborative filtering, matrix factorisation, content-based and embedding-based approaches, and two-stage retrieval and ranking. Chapters on evaluation cover offline metrics and online experiments, and later chapters tackle cold start, diversity, feedback loops and the ethics of optimising for engagement.
Worked examples use small datasets you can compute by hand.
What you will learn
- Baselines that are surprisingly hard to beat
- Collaborative filtering and matrix factorisation
- Content-based and embedding-based recommendation
- Two-stage retrieval and ranking architectures
- Offline metrics, A/B tests and what they miss
- Cold start, diversity, feedback loops and responsible design
Table of contents
- IntroductionIntroduction
- Chapter 1What Recommenders Do
- Chapter 2The Data: Users, Items and Interactions
- Chapter 3Explicit and Implicit Feedback
- Chapter 4Defining the Objective
- Chapter 5Baselines That Are Hard to Beat
- Chapter 6Co-occurrence and Association Rules
- Chapter 7Recency, Popularity and Time
- Chapter 8User-Based Collaborative Filtering
- Chapter 9Item-Based Collaborative Filtering
- Chapter 10Matrix Factorisation
- Chapter 11Content-Based Recommendation
- Chapter 12Embedding-Based Recommendation
- Chapter 13Hybrid Approaches
- Chapter 14Sequential and Session-Based Recommendation
- Chapter 15Context-Aware Recommendation
- Chapter 16Knowledge and Constraints
- Chapter 17The Two-Stage Architecture
- Chapter 18Candidate Generation
- Chapter 19Ranking Models
- Chapter 20Features for Ranking
- Chapter 21Position Bias
- Chapter 22Re-Ranking for Diversity and Rules
- Chapter 23Real-Time Serving
- Chapter 24Recommendation and Search
- Chapter 25Offline Evaluation
- Chapter 26NDCG, Worked Through
- Chapter 27Beyond Accuracy: Coverage, Novelty and Diversity
- Chapter 28Why Offline Metrics Mislead
- Chapter 29Online Experiments
- Chapter 30Interleaving
- Chapter 31Long-Term Effects
- Chapter 32Business Metrics
- Chapter 33Evaluation Checklist
- Chapter 34Logging: The Data Pipeline Behind Recommendations
- Chapter 35Matrix Factorisation, One Step at a Time
- Chapter 36Item-Based Recommendations in Code
- Chapter 37Recommendations in Email and Notifications
- Chapter 38Business-to-Business and Long Purchase Cycles
- Chapter 39Language Models in Recommender Systems
- Chapter 40Team, Process and Governance
- Chapter 41Approximate Nearest-Neighbour Search
- Chapter 42Debugging a Recommender
- Chapter 43Cold Start
- Chapter 44Exploration and Exploitation
- Chapter 45Feedback Loops and Filter Bubbles
- Chapter 46Fairness for Users and Providers
- Chapter 47Privacy
- Chapter 48Explanations
- Chapter 49User Control
- Chapter 50Content Quality and Safety
- Chapter 51Case Study: An Online Bookshop
- Chapter 52Case Study: A Streaming Service
- Chapter 53Case Study: A Job Board
- Chapter 54A Project Checklist
- AppendixExercises
- GlossaryGlossary
Who this book is for
Engineers, data scientists and product managers working on personalisation and recommendations.
Details
| Format | PDF + web edition |
|---|---|
| Length | 81 pages (approx.) |
| Chapters | 54 |
| Level | Intermediate |
| Language | English |
| Topic | AI Engineering |
| Author | Latent Library Editorial |
Keep reading