You are previewing Statistical Methods for Recommender Systems.
O'Reilly logo
Statistical Methods for Recommender Systems

Book Description

Designing algorithms to recommend items such as news articles and movies to users is a challenging task in numerous web applications. The crux of the problem is to rank items based on users' responses to different items to optimize for multiple objectives. Major technical challenges are high dimensional prediction with sparse data and constructing high dimensional sequential designs to collect data for user modeling and system design. This comprehensive treatment of the statistical issues that arise in recommender systems includes detailed, in-depth discussions of current state-of-the-art methods such as adaptive sequential designs (multi-armed bandit methods), bilinear random-effects models (matrix factorization) and scalable model fitting using modern computing paradigms like MapReduce. The authors draw upon their vast experience working with such large-scale systems at Yahoo! and LinkedIn, and bridge the gap between theory and practice by illustrating complex concepts with examples from applications they are directly involved with.

Table of Contents

  1. Cover
  2. Half title
  3. Dedication
  4. Title
  5. Copyright
  6. Table of Contents
  7. Preface
  8. Part I Introduction
    1. 1 Introduction
      1. 1.1 Overview of Recommender Systems for Web Applications
      2. 1.2 A Simple Scoring Model: Most-Popular Recommendation
      3. Exercises
    2. 2 Classical Methods
      1. 2.1 Item Characterization
      2. 2.2 User Characterization
      3. 2.3 Feature-Based Methods
      4. 2.4 Collaborative Filtering
      5. 2.5 Hybrid Methods
      6. 2.6 Summary
      7. Exercises
    3. 3 Explore-Exploit for Recommender Problems
      1. 3.1 Introduction to the Explore-Exploit Trade-off
      2. 3.2 Multiarmed Bandit Problem
      3. 3.3 Explore-Exploit in Recommender Systems
      4. 3.4 Explore-Exploit with Data Sparsity
      5. 3.5 Summary
      6. Exercise
    4. 4 Evaluation Methods
      1. 4.1 Traditional Offline Evaluation
      2. 4.2 Online Bucket Tests
      3. 4.3 Offline Simulation
      4. 4.4 Offline Replay
      5. 4.5 Summary
      6. Exercise
  9. Part II Common Problem Settings
    1. 5 Problem Settings and System Architecture
      1. 5.1 Problem Settings
      2. 5.2 System Architecture
    2. 6 Most-Popular Recommendation
      1. 6.1 Example Application: Yahoo! Today Module
      2. 6.2 Problem Definition
      3. 6.3 Bayesian Solution
      4. 6.4 Non-Bayesian Solutions
      5. 6.5 Empirical Evaluation
      6. 6.6 Large Content Pools
      7. 6.7 Summary
      8. Exercises
    3. 7 Personalization through Feature-Based Regression
      1. 7.1 Fast Online Bilinear Factor Model
      2. 7.2 Offline Training
      3. 7.3 Online Learning
      4. 7.4 Illustration on Yahoo! Data Sets
      5. 7.5 Summary
      6. Exercise
    4. 8 Personalization through Factor Models
      1. 8.1 Regression-Based Latent Factor Model (RLFM)
      2. 8.2 Fitting Algorithms
      3. 8.3 Illustration of Cold Start
      4. 8.4 Large-Scale Recommendation of Time-Sensitive Items
      5. 8.5 Illustration of Large-Scale Problems
      6. 8.6 Summary
      7. Exercise
  10. Part III Advanced Topics
    1. 9 Factorization through Latent Dirichlet Allocation
      1. 9.1 Introduction
      2. 9.2 Model
      3. 9.3 Training and Prediction
      4. 9.4 Experiments
      5. 9.5 Related Work
      6. 9.6 Summary
    2. 10 Context-Dependent Recommendation
      1. 10.1 Tensor Factorization Models
      2. 10.2 Hierarchical Shrinkage
      3. 10.3 Multifaceted News Article Recommendation
      4. 10.4 Related-Item Recommendation
      5. 10.5 Summary
    3. 11 Multiobjective Optimization
      1. 11.1 Application Setting
      2. 11.2 Segmented Approach
      3. 11.3 Personalized Approach
      4. 11.4 Approximation Methods
      5. 11.5 Experiments
      6. 11.6 Related Work
      7. 11.7 Summary
  11. Endnotes
  12. References
  13. Index