Recommender Systems Handbook

Preface
Contents
Recommender Systems: Techniques, Applications, and Challenges
	1 Introduction
	2 Recommender Systems' Value
	3 Data and Knowledge Sources
	4 Recommendation Techniques
	5 Special Recommendation Techniques
	6 Recommender Systems Evaluation
	7 Recommender Systems and Human-Computer Interaction
	8 Recommender Systems Applications
	9 Challenges
		9.1 Preference Acquisition
		9.2 Interaction
		9.3 New Recommendation Tasks
	References
Part I General Recommendation Techniques
	Trust Your Neighbors: A Comprehensive Survey of Neighborhood-Based Methods for Recommender Systems
		1 Introduction
			1.1 Advantages of Neighborhood Approaches
			1.2 Objectives and Outline
		2 Problem Definition and Notation
		3 Neighborhood-Based Recommendation
			3.1 User-Based Rating Prediction
			3.2 User-Based Classification
			3.3 Regression vs. Classification
			3.4 Item-Based Recommendation
			3.5 User-Based vs. Item-Based Recommendation
		4 Components of Neighborhood Methods
			4.1 Rating Normalization
				4.1.1 Mean-centering
				4.1.2 Z-score Normalization
				4.1.3 Choosing a Normalization Scheme
			4.2 Similarity Weight Computation
				4.2.1 Correlation-Based Similarity
				4.2.2 Other Similarity Measures
				4.2.3 Considering the Significance of Weights
				4.2.4 Considering the Variance of Ratings
				4.2.5 Considering the Target Item
			4.3 Neighborhood Selection
				4.3.1 Pre-filtering of Neighbors
				4.3.2 Neighbors in the Predictions
		5 Advanced Techniques
			5.1 Learning-Based Methods
				5.1.1 Factorization Methods
				5.1.2 Neighborhood-Learning Methods
			5.2 Graph-Based Methods
				5.2.1 Path-Based Similarity
				5.2.2 Random Walk Similarity
				5.2.3 Combining Random Walks and Neighborhood-Learning Methods
				5.2.4 User-Adaptive Diffusion Models
		6 Conclusion
		References
	Advances in Collaborative Filtering
		1 Introduction
		2 Preliminaries
			2.1 The Netflix Data
			2.2 Implicit Feedback as Auxiliary Data
		3 Baseline Predictors
			3.1 Time Changing Baseline Predictors
				3.1.1 Predicting Future Days
		4 Matrix Factorization Models
			4.1 Factorizing the Rating Matrix
			4.2 Matrix Factorization with Implicit Feedback
			4.3 Time-Aware Factor Model
			4.4 Comparison
			4.5 Summary
		5 Neighborhood Models
			5.1 Similarity Measures
			5.2 Similarity-Based Interpolation
			5.3 A Global Neighborhood Model
				5.3.1 Building the Model
				5.3.2 Parameter Estimation
				5.3.3 Comparison of Accuracy
			5.4 A Factorized Neighborhood Model
				5.4.1 Factoring Item-Item Relationships
				5.4.2 A User-User Model
			5.5 Temporal Dynamics at Neighborhood Models
			5.6 Summary
		6 Between Neighborhood and Factorization
		7 Factorization Machines
			7.1 Baseline Predictors Through Linear Regression
				7.1.1 Baseline with User and Item Biases
				7.1.2 Time Changing Baseline Predictors
				7.1.3 Global Neighborhood Model
			7.2 Factorization Machines for Collaborative Filtering
				7.2.1 Encoding SVD and Neighborhood Models
				7.2.2 Attribute-Aware ``Hybrid'' Models
			7.3 Learning Factorization Machines
			7.4 Comparison
			7.5 Discussion
		8 Beyond Rating Prediction for the Netflix Prize
			8.1 Recent Results for Rating Prediction
			8.2 Item Recommendation
			8.3 Generalizing to Other Recommendation Tasks
			8.4 Neural Networks
		References
	Item Recommendation from Implicit Feedback
		1 Introduction
		2 Problem Definition
			2.1 Recommender Modeling
				2.1.1 Dot Product Models
				2.1.2 Examples
			2.2 Applying Item Recommenders
			2.3 Evaluation
			2.4 Learning from Implicit Feedback
			2.5 Core Challenges of Item Recommendation
		3 Learning Objectives
			3.1 Pointwise Loss
			3.2 Pairwise Loss
			3.3 Softmax Loss
		4 Sampling Based Learning Algorithms
			4.1 Algorithms for Pointwise Loss
				4.1.1 Batching
				4.1.2 Omitted Details
			4.2 Algorithms for Pairwise Loss
				4.2.1 Uniform Sampling Without Weight
				4.2.2 Uniform Sampling with Weights
				4.2.3 Adaptive Sampling with Weights
				4.2.4 Adaptive Sampling Without Weights
			4.3 Algorithms for Sampled Softmax
				4.3.1 Kernel Based Sampling
				4.3.2 Two-Pass Sampler
				4.3.3 Relation of Sampled Softmax to Pairwise Loss
		5 Efficient Learning Algorithms for Special Cases
			5.1 Pointwise Square Loss
				5.1.1 Gramian Trick
				5.1.2 Coordinate Descent/ALS Solver for Multilinear Models
				5.1.3 SGD Solver for General Models
			5.2 Pairwise Square Loss
		6 Retrieval with Item Recommenders
			6.1 Limitations of Brute-Force Retrieval
			6.2 Approximate Nearest Neighbor Search
			6.3 Dynamic User Model
		7 Conclusion
		References
	Deep Learning for Recommender Systems
		1 Introduction
		2 Deep Learning for Recommender Systems: Preliminary
			2.1 Basics of Recommender Systems
			2.2 Basics of Deep Learning Techniques
				2.2.1 Multi-Layer Perceptrons
				2.2.2 Convolutional Neural Networks
				2.2.3 Recurrent Neural Networks
				2.2.4 Encoder and Decoder Architectures
				2.2.5 Graph Neural Networks
				2.2.6 Deep Reinforcement Learning
				2.2.7 Adversarial Neural Networks
				2.2.8 Restricted Boltzmann Machines
		3 Deep Learning for Recommender Systems: Algorithms
			3.1 Deep Learning for Interaction Modeling
				3.1.1 User-Item Interaction Modeling
				3.1.2 Feature Interaction Modeling
			3.2 Deep Learning for User Modeling
				3.2.1 Temporal Dynamics Modeling
				3.2.2 Diverse Interest Modeling
			3.3 Deep Learning for Content Representation Learning
				3.3.1 Textual Feature Extraction
				3.3.2 Image Feature Extraction
				3.3.3 Video and Audio Feature Extraction
			3.4 Deep Learning for Graph-Structured Data in Recommendation
			3.5 Deep Learning for Cold-Start Recommendation
			3.6 Deep Learning for Recommendation: Beyond Accuracy
				3.6.1 Explainable Recommendations
				3.6.2 Robust Recommender Systems
			3.7 Deep Reinforcement Learning for Recommendation
		4 Deep Learning for Recommender Systems: Applications
			4.1 Deep Learning for E-commerce Recommendation
			4.2 Deep Learning for Online Entertainment Recommendation
			4.3 Deep Learning Based News Recommendation
			4.4 Deep Learning for Point-of-Interest Recommendation
			4.5 Deep Learning Based Recommendation on Other Domains
		5 Discussion and Conclusion
		References
	Context-Aware Recommender Systems: From Foundations to Recent Developments
		1 Introduction and Motivation
		2 Context in Recommender Systems
			2.1 What Is Context?
			2.2 Modeling Contextual Information in RS: Traditional Approach
			2.3 Modeling Contextual Information in RS: Major Approaches
			2.4 Designing and Obtaining Contextual Factors
		3 Paradigms for Incorporating Context in RS
		4 Contextual Modeling Approaches
		5 Discussion and Conclusions
		References
	Semantics and Content-Based Recommendations
		1 Introduction
		2 Content-Based Recommender Systems
			2.1 The Architecture of a Content-Based Recommender System
			2.2 Semantics-Aware Content Representation
				2.2.1 Endogenous Semantics
				2.2.2 Exogenous Semantics
			2.3 Strengths and Weaknesses of Content-Based Recommendations
		3 Recent Developments and New Trends
			3.1 Embeddings and Distributed Representations
				3.1.1 Recommender Systems Based on Word and Sentence Embeddings
				3.1.2 Deep Learning Models Based on Word and Sentence Embeddings
			3.2 Linked Open Data and Knowledge Graphs
			3.3 User-Generated Content and Multimedia Features
				3.3.1 Content-Based Recommender Systems Leveraging User-Generated Content
				3.3.2 Content-Based Recommender Systems Leveraging Multimedia Content
			3.4 Transparency and Content-Based Explanations
				3.4.1 Generating Explainable Content-Based Recommendations
				3.4.2 Generating Post Hoc Content-Based Explanations
			3.5 Exploiting Content for Conversational Recommender Systems
				3.5.1 Dialog State-Based CRS
				3.5.2 End-to-End Systems
		4 Discussion and Future Outlook
		5 Conclusions
		References
Part II Special Recommendation Techniques
	Session-Based Recommender Systems
		1 Introduction
		2 Problem Characterization
		3 Technical Approaches
			3.1 Model-Free Approaches
				3.1.1 Sequential Pattern Mining
				3.1.2 Nearest Neighbors
			3.2 Model-Based Approaches
				3.2.1 Markov Models
				3.2.2 Recurrent Neural Networks
				3.2.3 Reinforcement Learning
		4 Evaluation of Session-Based Recommenders
			4.1 Offline Evaluation Protocols
				4.1.1 Dataset Partitioning
				4.1.2 Definition of Target Interactions
				4.1.3 Evaluation Metrics
			4.2 Evaluation of Reinforcement Learning Approaches
			4.3 Datasets
			4.4 User-Centric Evaluation of Session-Based Recommendations
			4.5 Limitations of Current Research
				4.5.1 Methodological Issues and Limitations of Academic Research
				4.5.2 Potential Biases in Datasets
		5 Future Directions
			5.1 Algorithmic Improvements: Considering More Types of Data
			5.2 Improving the Evaluation Methodology
		References
	Adversarial Recommender Systems: Attack, Defense, and Advances
		1 Introduction
		2 Foundations
			2.1 Adversarial Perspective
			2.2 Taxonomy of Adversarial Attacks
				2.2.1 Attack's Timing
				2.2.2 Attacker's Knowledge
				2.2.3 Attacker's Goals
			2.3 Adversarial Robustness: A Unified View of Adversarial Attacks and Defenses
			2.4 Definition of Adversarial Attack and Countermeasure Strategies
				2.4.1 Attack Models
				2.4.2 Adversarial Countermeasure Strategies
		3 A Classification of Adversarial Attacks on Defenses of RS
			3.1 Adversarial Perturbation of Model Embedding Parameters
			3.2 Adversarial Perturbations on Content Data
			3.3 Machine-Learned Data Poisoning Attacks
				3.3.1 Data Poisoning Optimization on Factorization-Based Recommenders
				3.3.2 Data Poisoning Optimization and Reinforcement Learning
				3.3.3 Data Poisoning Optimization Poisoning with Other Recommendation Families
		4 Evaluation
			4.1 The Experimental Setting
			4.2 Evaluation Metrics
				4.2.1 Impact on Overall Recommendation Performance
				4.2.2 Impact on the Recommendability of Item Categories
				4.2.3 Qualitative Evaluation of Perturbed Content
		5 Conclusion and the Road Ahead
		References
	Group Recommender Systems: Beyond Preference Aggregation
		1 Introduction
		2 Usage Scenarios and Classification of Group Recommenders
			2.1 Usage Scenario 1: Interactive Television
			2.2 Usage Scenario 2: Ambient Intelligence
			2.3 Usage Scenarios Underlying Related Work
			2.4 A Classification of Group Recommenders
		3 Aggregation Strategies
			3.1 Overview of Aggregation Strategies
			3.2 Aggregation Strategies Used in Related Work
			3.3 Which Strategy Performs the Best
			3.4 Evaluating Group Recommender Systems
				3.4.1 User Studies
				3.4.2 Off-Line Evaluations
		4 Impact of Sequence Order
		5 Modeling Affective State
			5.1 Effects of the Group on an Individual's Satisfaction
		6 Using Satisfaction Inside Aggregation Strategies
		7 Incorporating Group Attributes
		8 Applying Group Recommendation to Individual Users
			8.1 Multiple Criteria
			8.2 Cold-Start Problem
			8.3 Virtual Group Members
		9 Conclusions and Challenges
			9.1 Main Issues Raised
			9.2 Caveat: Group Modeling
			9.3 Challenges
		References
	People-to-People Reciprocal Recommenders
		1 Introduction
		2 Reciprocal vs. Traditional Recommenders
		3 Previous Work
			3.1 Social Networks
			3.2 Mentor-Mentee Matching
			3.3 Online Learning Courses
			3.4 Job Recommendation
			3.5 Online Dating
		4 A Case Study in Online Dating
			4.1 CCR: Content-Collaborative Reciprocal Recommender for Online Dating
				4.1.1 Evaluation
			4.2 Explicit and Implicit User Preferences
				4.2.1 Are Explicit Preferences  Good Predictors of User Interactions?
				4.2.2 Are Implicit Preferences  Good Predictors of User Interactions?
		5 Conclusions
		6 Challenges and Future Directions
		References
	Natural Language Processing for Recommender Systems
		1 Introduction
		2 User Generated Reviews
			2.1 Affinity to Sentiment Analysis
			2.2 Traditional Methods
			2.3 Deep Learning: Preliminaries
				2.3.1 Text Processor
				2.3.2 Machine Learning Annotation
			2.4 Review-Based Recommenders for Rating Prediction
				2.4.1 DeepCoNN
				2.4.2 TransNets
				2.4.3 Extended TransNets
			2.5 State-of-the-Art of Review-Based Recommendations
				2.5.1 Motivation
				2.5.2 Intuition
				2.5.3 Algorithm Overview
				2.5.4 Phase I: Distribution of Matching Vectors
				2.5.5 Phase II: Collaborative Filtering with Augmented Labels
			2.6 Empirical Evaluation
			2.7 Review-Based Recommenders for Ranking
				2.7.1 Input Source Modeling
				2.7.2 Integrated Representation
				2.7.3 Ranking Optimization
				2.7.4 In-Depth Analysis
				2.7.5 Empirical Evaluation
			2.8 Discussion and Future Outlook
		3 Conversational Preference Elicitation
			3.1 Conversational Recommender Systems
			3.2 Critiquing
				3.2.1 CE-NCF
				3.2.2 Latent Linear Conversational Critiquing
			3.3 Facets-Based Preference Elicitation
			3.4 Question-Based Preference Elicitation
				3.4.1 ``System Ask, User Respond'' (SAUR)
				3.4.2 Topic-Based Questions
				3.4.3 Asking Absolute vs. Relative Questions
			3.5 Discussion and Future Outlook
		4 Generating Textual Explanations
			4.1 Review-Level Explanations
				4.1.1 Multi-Task Learning for Recommendation and Personalized Explanation
				4.1.2 Providing Explanations for Recommendations in Reciprocal Environments
			4.2 Feature-Level Explanations
				4.2.1 Explicit Factor Models for Explainable Recommendation Based on Phrase-Level Sentiment Analysis
				4.2.2 Social Collaborative Viewpoint Regression with Explainable Recommendations
				4.2.3 Review-Aware Explainable Recommendation by Modeling Aspects
			4.3 Discussion and Future Outlook
		References
	Design and Evaluation of Cross-Domain Recommender Systems
		1 Introduction
		2 Formulation of the Cross-Domain Recommendation Problem
			2.1 Definition of Domain
			2.2 Cross-Domain Recommendation Tasks
			2.3 Cross-Domain Recommendation Goals
			2.4 Cross-Domain Recommendation Scenarios
		3 Categorization of Cross-Domain Recommendation Techniques
		4 Cross-Domain Recommendations Techniques
			4.1 Merging User Preferences
			4.2 Linking Domains
			4.3 Transfer Learning
			4.4 Co-Training of Shared Latent Features
			4.5 Deep Learning
				4.5.1 Transfer Learning
				4.5.2 Co-Training
		5 Evaluation of Cross-Domain Recommender Systems
			5.1 Data Partitioning
			5.2 Sensitivity Analysis
		6 Open Research Questions
		7 Conclusions
		References
Part III Value and Impact of Recommender Systems
	Value and Impact of Recommender Systems
		1 Introduction
		2 Stakeholders and Value Drivers of Recommender Systems
			2.1 Stakeholders of Recommender Systems
			2.2 Value Dimensions of Recommender Systems
				2.2.1 Efficiency
				2.2.2 Complementarities
				2.2.3 Lock-In
				2.2.4 Novelty
			2.3 Risks of Recommender Systems
		3 Measuring the Impact of Recommender Systems
			3.1 Value Dimensions and Measurements in Practical Applications
			3.2 Click-Through-Rate
			3.3 Adoption and Conversion Measures
			3.4 Sales and Revenue
			3.5 User Engagement
			3.6 Effects on Consumption Distributions
		4 Towards More Value-Oriented and Impactful Research
			4.1 Recommender Systems Research with a Purpose
			4.2 Utilizing a Richer Methodological Repertoire
				4.2.1 Evaluating with Humans in the Loop
				4.2.2 Re-thinking Data-Based Research
		5 Conclusions
		References
	Evaluating Recommender Systems
		1 Introduction
		2 Experimental Settings
			2.1 Offline Experiments
				2.1.1 Data Sets for Offline Experiments
				2.1.2 Simulating User Behavior
				2.1.3 More Complex User Modeling
			2.2 User Studies
				2.2.1 Advantages and Disadvantages
				2.2.2 Between vs. Within Subjects
				2.2.3 Variable Counter Balance
				2.2.4 Questionnaires
			2.3 Online Evaluation
			2.4 Offline-Online Correlations
			2.5 Drawing Reliable Conclusions
				2.5.1 Confidence and p-values
				2.5.2 Paired Results
				2.5.3 Unpaired Results
				2.5.4 Multiple Tests
				2.5.5 Confidence Intervals
			2.6 Reporting Results
				2.6.1 Reporting the Experimental Settings
				2.6.2 Compared Methods
				2.6.3 Reporting Complete Results
			2.7 Standardized Evaluation Frameworks
		3 Evaluation in the Industry
		4 Recommender System Properties
			4.1 User Preference
			4.2 Prediction Accuracy
				4.2.1 Measuring Ratings Prediction Accuracy
				4.2.2 Measuring Usage Prediction
				4.2.3 Ranking Measures
			4.3 Coverage
				4.3.1 Item Space Coverage
				4.3.2 User Space Coverage
				4.3.3 Cold Start
			4.4 Confidence
			4.5 Trust
			4.6 Novelty
			4.7 Serendipity
			4.8 Diversity
			4.9 Utility
			4.10 Risk
			4.11 Robustness
			4.12 Privacy
			4.13 Adaptivity
			4.14 Scalability
		5 Conclusion
		References
	Novelty and Diversity in Recommender Systems
		1 Introduction
		2 Novelty and Diversity in Recommender Systems
			2.1 Why Novelty and Diversity in Recommendation
				2.1.1 System Perspective
				2.1.2 User Perspective
				2.1.3 Business Perspective
				2.1.4 The Limits of Novelty and Diversity
			2.2 Defining Novelty and Diversity
			2.3 Diversity in Other Fields
		3 Novelty and Diversity Evaluation
			3.1 Notation
			3.2 Average Intra-List Distance
			3.3 Global Long-Tail Novelty
			3.4 User-Specific Unexpectedness
			3.5 Inter-Recommendation Diversity Metrics
			3.6 Specific Methodologies
			3.7 Diversity vs. Novelty vs. Serendipity
			3.8 Information Retrieval Diversity
			3.9 Proportional Diversity
		4 Novelty and Diversity Enhancement Approaches
			4.1 Result Diversification/Re-ranking
			4.2 Using Clustering for Diversification
			4.3 Fusion-Based Methods
			4.4 Incorporating Diversity in the Ranking Objective
			4.5 Serendipity: Enabling Surprising Recommendations
			4.6 Other Approaches
			4.7 User Studies
			4.8 Diversification Approaches in Information Retrieval
		5 Unified View
			5.1 General Novelty/Diversity Metric Scheme
			5.2 Item Novelty Models
				5.2.1 Item Discovery
				5.2.2 Item Familiarity
			5.3 Resulting Metrics
				5.3.1 Discovery-Based
				5.3.2 Familiarity-Based
				5.3.3 Further Unification
				5.3.4 Direct Optimization of Novelty Models
			5.4 Connecting Recommendation Diversity and Search Diversity
				5.4.1 Rank and Relevance
				5.4.2 IR Diversity in Recommendation
				5.4.3 Personalized Diversity
		6 Bias and Fairness
			6.1 Bias in Recommendation
			6.2 Fair Recommendation
		7 Empirical Metric Comparison
		8 Conclusion
		References
	Multistakeholder Recommender Systems
		1 Introduction
		2 Recommendation Stakeholders and the Multistakeholder Paradigm
			2.1 Multistakeholder Evaluation
			2.2 Multistakeholder Algorithms
			2.3 Multistakeholder Design
			2.4 The Landscape of Multistakeholder Recommendation
				2.4.1 Consumer-Side Issues
				2.4.2 Provider-Side Issues
				2.4.3 System/Platform Issues
				2.4.4 Side Stakeholders
				2.4.5 Tensions Among Stakeholders
		3 Related Research
			3.1 Economic Foundations
			3.2 Multi-Objective Recommendation
			3.3 Personalization for Matching
			3.4 Group Recommendation
		4 Evaluation
			4.1 Simulation
			4.2 Models of Utility
			4.3 Off-Line Experiment Design
			4.4 User Studies
			4.5 Metrics
				4.5.1 Provider Metrics
				4.5.2 System Metrics
		5 Multistakeholder Evaluation Example: Popularity Bias in Recommendation
			5.1 Data
			5.2 Algorithmic Bias
			5.3 Multistakeholder Impact
				5.3.1 Item Groups
				5.3.2 User Groups
				5.3.3 Provider Groups
			5.4 Summary
		6 Conclusion and Future Work
		References
	Fairness in Recommender Systems
		1 Introduction
			1.1 Example Applications
			1.2 Fundamental Concepts in (Un)Fairness
			1.3 Multisided Analysis of Recommendation
			1.4 Unfairness in Recommendation
		2 Studying Fairness
			2.1 Defining Fairness
			2.2 Collecting Data
			2.3 Structuring Experiments
			2.4 Reporting Results
		3 Consumer Fairness
			3.1 Individual Fairness
			3.2 Group Fairness
			3.3 Fairness Beyond Accuracy
			3.4 More Complex Scenarios
		4 Provider Fairness
			4.1 Provider Utility
			4.2 Individual Fairness
			4.3 Group Fairness
		5 Other Stakeholders
		6 Fairness over Time
		7 Methods for Mitigation
			7.1 Fairness in Data
			7.2 Fairness in Ranking Models
			7.3 Fairness Through Re-ranking
			7.4 Fairness Through Engineering
		8 Conclusion
		References
Part IV Human Computer Interaction
	Beyond Explaining Single Item Recommendations
		1 Introduction
		2 Presentation of, and Interaction with, Recommendations
			2.1 Presenting Recommendations
			2.2 Preference Elicitation
			2.3 The User Specifies Their Requirements
			2.4 The User Asks for an Alteration
			2.5 The User Rates Items
			2.6 The User Gives Their Opinion
			2.7 Mixed Interaction Interfaces
			2.8 Interactive Recommendations
		3 Explanation Levels and Styles
			3.1 Level 1: Individual User
				3.1.1 Content-Based Style Explanation
				3.1.2 Case-Based Reasoning (CBR) Style Explanations
				3.1.3 Knowledge and Utility-Based Style Explanations
			3.2 Level 2: Contextualization
				3.2.1 Preference Space
				3.2.2 Demographic-Based Style Explanations
				3.2.3 Collaborative-Based Style Explanations
				3.2.4 Interactive Contextualization
			3.3 Level 3: Self-Actualization
			3.4 Hybrid Style Explanations
		4 Goals and Metrics
			4.1 Explain How the System Works: Transparency
			4.2 Allow Users to Tell the System It Is Wrong: Scrutability
			4.3 Increase Users' Confidence in the System: Trust
			4.4 Convince Users to Try or Buy: Persuasiveness
			4.5 Help Users Make Good Decisions: Effectiveness
			4.6 Help Users Make Decisions Faster: Efficiency
			4.7 Make the Use of the System Enjoyable: Satisfaction
			4.8 Additional Goals
		5 Moderating Factors
			5.1 Personal Characteristics
			5.2 Situational Characteristics
		6 Future Directions
			6.1 Generating from Reviews
			6.2 Social Recommendations
			6.3 Explanations to Deal with Bias
			6.4 Sets and Sequences
			6.5 Over- and Under Reliance
		References
	Personality and Recommender Systems
		1 Introduction
		2 Personality Model
			2.1 The Five Factor Model of Personality
			2.2 Other Models of Personality
			2.3 Relationship Between Personality and User Preferences
		3 Personality Acquisition
			3.1 Explicit Personality Acquisition
			3.2 Implicit Personality Acquisition
			3.3 Datasets for Offline Recommender Systems Experiments
		4 How to Use Personality in Recommender Systems
			4.1 Addressing the New User Problem
			4.2 Personalizing Recommendation Diversity
			4.3 Other Applications
		5 Open Issues and Challenges
			5.1 Non-intrusive Acquisition of Personality Information
			5.2 Larger Datasets
			5.3 Cross-Domain Applications
			5.4 Beyond Accuracy
			5.5 Privacy Issues
		6 Conclusion
		References
	Individual and Group Decision Making and Recommender Systems
		1 Introduction and Preview
			1.1 Choice Support as a Central Goal
			1.2 Different Degrees of Human Involvement
			1.3 Preview of This Chapter
			1.4 What Constitutes a Good Choice for an Individual?
		2 Choice Patterns and Recommendation to Individuals
			2.1 Attribute-Based Choice
			2.2 Consequence-Based Choice
			2.3 Socially Based Choice
			2.4 Experience-Based Choice
			2.5 Policy-Based Choice
			2.6 Trial-and-Error-Based Choice
			2.7 Combinations of Choice Patterns
		3 Choice Support Strategies and Recommendation to Individuals
			3.1 Evaluate on Behalf of the Chooser
			3.2 Advise About Processing
			3.3 Access Information and Experience
			3.4 Represent the Choice Situation
			3.5 Combine and Compute
			3.6 Design the Domain
			3.7 Support Communication
		4 New Considerations With Regard to Groups
			4.1 Relevant Areas of Research and Practice
			4.2 The Importance of Differences Among Group Members
			4.3 What Constitutes a Good Choice for a Group?
			4.4 Principles from Research and Practice
			4.5 High-Level Approaches for Group Recommender Systems
				4.5.1 Approach 1: Support Interaction
				4.5.2 Approach 2: Predict the Results of Interaction
		5 Choice Patterns and Recommendation to Groups
			5.1 Attribute-Based Choice
			5.2 Consequence-Based Choice
			5.3 Socially Based Choice
			5.4 Experience-Based Choice
			5.5 Policy-Based Choice
			5.6 Trial-and-Error-Based Choice
			5.7 Combinations of Choice Patterns
		6 Choice Support Strategies and Recommendation to Groups
			6.1 Evaluate on Behalf of the Chooser
			6.2 Advise About Processing
			6.3 Access Information and Experience
			6.4 Represent the Choice Situation
			6.5 Combine and Compute
			6.6 Design the Domain
			6.7 Support Communication
		7 Recapitulation and Concluding Remarks
		References
Part V Recommender Systems Applications
	Social Recommender Systems
		1 Introduction
		2 Content Recommendation
			2.1 Key Domains
			2.2 Group Recommender Systems: Beyond Preference Aggregation
			2.3 Case Study: Social Media Recommendation in the Enterprise
			2.4 Summary
		3 People Recommendation
			3.1 Recommending People to Connect With
			3.2 Recommending Strangers
			3.3 Recommending People to Follow
			3.4 Related Research Areas
			3.5 Summary
		4 Discussion
		5 Emerging Domains and Open Challenges
			5.1 Emerging Domains
			5.2 Open Challenges
		References
	Food Recommender Systems
		1 Introduction to the Food Recommendation  Problem
		2 Problem Description
			2.1 User Roles and Groups
			2.2 Item Types
			2.3 Types of Food Choices
			2.4 Cooking Recommender
			2.5 Grocery Recommender
			2.6 Restaurant Recommender
			2.7 Health Recommender
		3 Algorithms
			3.1 Food Recommendation as a User-Item Ranking Problem
			3.2 Context-Dependent Food Recommendations
			3.3 Preferences Vary Across User Groups
			3.4 Variations on the Food Recommendation Problem
		4 Interfaces
			4.1 Presenting and Accessing Recommendations
			4.2 Eliciting User Needs and Preferences
			4.3 Commercial Applications
		5 Evaluation
		6 Implementation Resources
			6.1 Recipe Datasets
			6.2 Grocery Datasets
			6.3 Meals, Menus and Restaurant Datasets
			6.4 Flavour Resources
			6.5 Software Frameworks
			6.6 Nutrition Resources
			6.7 Health Resources
			6.8 Food Sustainability Resources
			6.9 Other Resources
		7 Conclusion
		References
	Music Recommendation Systems: Techniques, Use Cases, and Challenges
		1 Introduction
			1.1 Characteristics of the Music Recommendation Domain
			1.2 Scope and Structure of the Chapter
		2 Types of Use Cases
			2.1 Basic Music Recommendation
				2.1.1 Interaction and Feedback Data
				2.1.2 Evaluation Metrics and Competitions
			2.2 Lean-in Exploration
				2.2.1 Evaluation Metrics and Competitions
				2.2.2 Discussion
			2.3 Lean-Back Listening
				2.3.1 Lean-Back Data and Evaluation
				2.3.2 Discussion
			2.4 Other Applications
		3 Types of Music Recommender Systems
			3.1 Collaborative Filtering
			3.2 Content-Based Filtering
			3.3 Hybrid Approaches
			3.4 Context-Aware Approaches
			3.5 Sequential Recommendation
			3.6 Psychology-Inspired Approaches
		4 Challenges
			4.1 How to Ensure and Measure Multi-Faceted Qualities of Recommendation Lists?
				4.1.1 Similarity Versus Diversity
				4.1.2 Novelty Versus Familiarity
				4.1.3 Popularity, Hotness, and Trendiness
				4.1.4 Serendipity
				4.1.5 Sequential Coherence
			4.2 How to Consider Intrinsic User Characteristics?
			4.3 How to Make a Music Recommender Fair?
			4.4 How to Explain Recommendations?
			4.5 How to Evaluate a Music Recommender System?
				4.5.1 Offline Evaluation
				4.5.2 Online Evaluation
				4.5.3 User Studies
			4.6 How to Deal with Missing and Negative Feedback in Evaluation?
			4.7 How to Design User Interfaces That Match the Use Case and Increase User Experience?
			4.8 Which Open Tools and Data Sources can be Used to Build a Music Recommender System?
		5 Conclusions
		References
	Multimedia Recommender Systems: Algorithms and Challenges
		1 Introduction
		2 Key Challenges
		3 Fundamentals of Feature Extraction and Item Representation
			3.1 Image Representation
			3.2 Video Representation
			3.3 Audio Representation
		4 Recommendation by Incorporating Multimedia Side Information
			4.1 Traditional Hybrid Approaches
				4.1.1 Extending Memory-Based Collaborative Filtering
				4.1.2 Extending Model-Based Collaborative Filtering
			4.2 Neural Approaches
				4.2.1 Answers to Challenges
			4.3 Graph-Based Approaches
				4.3.1 Random Walk-Based Method
				4.3.2 Graph Convolutional Network-Based Method
				4.3.3 Graph Autoencoder-Based Method
				4.3.4 Answers to Challenges
		5 Conclusions and Open Challenges
		References
	Fashion Recommender Systems
		1 Introduction
		2 Single Fashion Item Recommendations
			2.1 Similar Fashion Item Recommendations
				2.1.1 Collaborative Filtering  Similar Fashion Items Recommendations
				2.1.2 Content-Based Similar Fashion Items Recommendations
			2.2 Complementary Clothing  Item Recommendations
			2.3 Evaluating Single Item Fashion Recommendations
				2.3.1 Offline Evaluation of Single Item Fashion Recommendations
				2.3.2 Online Evaluation of Single Item Fashion Recommendations
				2.3.3 Beyond Accuracy Evaluation of Single Item Fashion Recommendations
		3 Social Fashion  Recommendations
			3.1 Leveraging Social Networking for Fashion Recommendation
			3.2 Methods for Social Fashion Recommendations
				3.2.1 Embeddings for Fashion Recommendations
				3.2.2 Social Influence  for Fashion Recommendations
			3.3 Evaluating Social Fashion Recommendations
		4 Size and Fit in Fashion
			4.1 Evaluating Size and Fit Recommendations
		5 Outfit Recommendations
			5.1 Outfits' Compatibility Scoring
			5.2 Sequential Outfits Representations and Predictors
			5.3 Outfits' Flexible Representations
			5.4 Evaluating Outfits Recommendations
		6 Fashion Recommendation Datasets
		7 Conclusion and Challenges Ahead
		References
Index