Building a Machine Learning Recommendation System
Chapter 1: Foundations of Recommendation Systems
What are Recommendation Systems?
Taxonomy of Recommender Engines
Understanding User-Item Interaction Data
Implicit vs. Explicit Feedback
The Cold-Start Problem
Preparing Your Development Environment
Practice: Loading and Inspecting a Dataset
Chapter 2: Content-Based Filtering
The Mechanics of Content-Based Recommenders
Creating Item Profiles from Metadata
Vectorizing Text Data with TF-IDF
Computing Similarity with Cosine Distance
Generating User Profiles
Producing Content-Based Recommendations
Hands-on Practical: Building a Movie Recommender
Chapter 3: Neighborhood-Based Collaborative Filtering
The Rationale Behind Collaborative Filtering
The User-Item Interaction Matrix
User-Based vs. Item-Based Approaches
Finding Similar Users and Items with k-Nearest Neighbors
Calculating Similarity Metrics
Making Predictions with Weighted Averages
Addressing Sparsity in the Interaction Matrix
Hands-on Practical: Implementing an Item-Based Filter
Chapter 4: Model-Based Collaborative Filtering with Matrix Factorization
From Neighborhoods to Latent Factor Models
Introduction to Matrix Factorization
Singular Value Decomposition (SVD) for Recommendations
Optimization with Stochastic Gradient Descent (SGD)
Regularization to Prevent Overfitting
Using Libraries for Matrix Factorization
Hands-on Practical: Generating Recommendations with SVD
Chapter 5: Evaluating Recommendation Systems
The Importance of Recommender Evaluation
Offline vs. Online Evaluation Methods
Splitting Data for Recommender Evaluation
Prediction Accuracy Metrics: RMSE and MAE
Ranking Metrics: Precision and Recall at K
Mean Average Precision (MAP)
Normalized Discounted Cumulative Gain (NDCG)
Hands-on Practical: Measuring Model Performance
Chapter 6: Constructing a Hybrid Recommendation System
Limitations of Single-Algorithm Systems
Combining Recommender Models
Weighted Hybridization
Switching and Mixed Hybridization Techniques
Feature Combination Methods
Building a System that Blends Content and Collaborative Signals
Hands-on Practical: A Weighted Hybrid Movie Recommender
Singular Value Decomposition (SVD) for Recommendations
Was this section helpful?
- Factorization Meets the Neighborhood: a Multifaceted Collaborative Filtering Model, Yehuda Koren, Robert Bell, and Chris Volinsky, 2008 Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (ACM) DOI: 10.1145/1401890.1401944 - Describes matrix factorization techniques, including SVD-like methods, for collaborative filtering and addressing sparse data, as developed for the Netflix Prize.
- Recommender Systems Handbook, Francesco Ricci, Lior Rokach, and Bracha Shapira, 2022 (Springer) DOI: 10.1007/978-1-0716-1738-4 - Provides a comprehensive overview of recommender systems, with dedicated chapters on matrix factorization and SVD variants.
- Mining of Massive Datasets, Jure Leskovec, Anand Rajaraman, Jeff Ullman, 2020 (Cambridge University Press) - Discusses SVD as a method for dimensionality reduction and its application in recommender systems, providing practical examples.