Liu / Datta / Lim Computational Trust Models and Machine Learning

1. Auflage 2014
ISBN: 978-1-4822-2667-6
Verlag: Taylor & Francis
Format: PDF
Kopierschutz: Adobe DRM (»Systemvoraussetzungen)

E-Book, Englisch, 232 Seiten

Computational Trust Models and Machine Learning provides a detailed introduction to the concept of trust and its application in various computer science areas, including multi-agent systems, online social networks, and communication systems. Identifying trust modeling challenges that cannot be addressed by traditional approaches, this book:

- Explains how reputation-based systems are used to determine trust in diverse online communities

- Describes how machine learning techniques are employed to build robust reputation systems

- Explores two distinctive approaches to determining credibility of resources—one where the human role is implicit, and one that leverages human input explicitly

- Shows how decision support can be facilitated by computational trust models

- Discusses collaborative filtering-based trust aware recommendation systems

- Defines a framework for translating a trust modeling problem into a learning problem

- Investigates the objectivity of human feedback, emphasizing the need to filter out outlying opinions

Computational Trust Models and Machine Learning effectively demonstrates how novel machine learning techniques can improve the accuracy of trust assessment.

Zielgruppe

Graduate students and researchers of machine learning and data mining techniques.

Autoren/Hrsg.

Liu, Xin

Datta, Anwitaman

Lim, Ee-Peng

Fachgebiete

• Mathematik | Informatik EDV | Informatik Informatik Theoretische Informatik

Inhaltsverzeichnis

Preface

List of Figures

List of Tables

Contributors

Introduction

Overview

What is Trust?

Computational Trust

Computational Trust Modeling: A Review

Machine Learning for Trust Modeling

Structure of the Book

Trust in Online Communities

Introduction

Trust in E-Commerce Environments

Trust in Search Engines

Trust in P2P Information Sharing Networks

Trust in Service-Oriented Environments

Trust in Social Networks

Discussion

Judging the Veracity of Claims and Reliability of Sources with Fact-Finders

Introduction

Related Work

Foundations of Trust

Consistency in Information Extraction

Source Dependence

Comparison to Other Trust Mechanisms

Fact-Finding

Priors

Fact-Finding Algorithms

Generalized Constrained Fact-Finding

Generalized Fact-Finding

Rewriting Fact-Finders for Assertion Weights

Encoding Information in Weighted Assertions

Encoding Groups and Attributes as Layers of Graph Nodes

Constrained Fact-Finding

Propositional Linear Programming

The Cost Function

Values ! Votes ! Belief

LP Decomposition

Tie Breaking

"Unknown" Augmentation

Experimental Results

Data

Experimental Setup

Generalized Fact-Finding

Constrained Fact-Finding

The Joint Generalized Constrained Fact-Finding Framework

Conclusion

Web Credibility Assessment

Introduction

Web Credibility Overview

What Is Web Credibility?

Introduction to Research on Credibility

Current Research

Definitions Used in This Chapter

Data Collection

Collection Means

Supporting Web Credibility Evaluation

Reconcile - A Case Study

Analysis of Content Credibility Evaluations

Subjectivity

Consensus and Controversy

Cognitive Bias

Aggregation Methods: What Is the Overall Credibility?

How to Measure Credibility

Standard Aggregates

Combating Bias: Whose Vote Should Count More?

Classifying Credibility Evaluations Using External Web Content Features

How We Get Values of Outcome Variables

The Motivation for Building a Feature-Based Classifier of Web Pages Credibility

Classification of Web Pages Credibility: Related Work

Dealing with Problem of Controversy

Aggregation of Evaluations

Features

The Results of Experiments with Build of Classifier Determining Whether a Web Page is Highly Credible (HC), Neutral (N) or Highly Not Credible (HNC)

The Results of Experiments with Build of Binary Classifier Determining Whether Webpage is Credible or Not

The Results of Experiments with Build of Binary Classifier of Controversy

Summary and Improvement Suggestions

Trust-Aware Recommender Systems

Recommender Systems

Content-Based Recommendation

Collaborative Filtering (CF)

Hybrid Recommendation

Evaluating Recommender Systems

Challenges of Recommender Systems

Summary

Computational Models of Trust in Recommender Systems

Definition and Properties

Global and Local Trust Metrics

Inferring Trust Values

Summary

Incorporating Trust in Recommender Systems

Trust-Aware Memory-Based CF Systems

Trust-Aware Model-Based CF Systems

Recommendation Using Distrust Information

Advantages of Trust-Aware Recommendation

Research Directions of Trust-Aware Recommendation

Conclusion

Biases in Trust-Based Systems

Introduction

Types of Biases

Cognitive Bias

Spam

Detection of Biases

Unsupervised Approaches

Supervised Approaches

Lessening the Impact of Biases

Avoidance

Aggregation

Compensation

Elimination

Summary

Glossary

Bibliography

Index

Über Autor(innen)

Xin Liu is currently a postdoctoral researcher in the Laboratoire de Systèmes d'Informations Répartis, led by Professor Karl Aberer, at École Polytechnique Fédérale de Lausanne (EPFL), Switzerland. Before joining EPFL, Xin received his Ph.D in computer science from Nanyang Technological University in Singapore, supervised by Associate Professor Anwitaman Datta. His current research interests include recommender systems, trust and reputation systems, social computing, and distributed computing. His papers have been accepted at several prestigious academic events, and he has been a program committee member and reviewer for numerous international conferences and journals.

Anwitaman Datta is an associate professor at Nanyang Technological University, Singapore, where he leads the Self-* Aspects of Networked and Distributed Systems Research Group and teaches courses on security management and cryptography and network security. Well published, he has focused his research on P2P storage, decentralized online social networks, structured overlays, and computational trust. His current research interests include the design of resilient large-scale distributed systems, coding for storage, security and privacy, and social media analysis. His projects have been funded by the Singapore Ministry of Education, HP Labs Innovation Research Award, and more.

Ee-Peng Lim is a professor at Singapore Management University (SMU), co-director of the SMU/Carnegie Mellon University Living Analytics Research Center, and associate editor of numerous journals and publications. He holds a Ph.D from the University of Minnesota, Minneapolis, USA and a B.Sc from the National University of Singapore. His current research interests include social network and web mining, information integration, and digital libraries. A former ACM Publications Board member, he currently serves on the steering committees of the International Conference on Asian Digital Libraries, Pacific Asia Conference on Knowledge Discovery and Data Mining, and International Conference on Social Informatics.