Collective view prediction is to judge the opinions of an active web
user based on unknown elements by referring to the collective mind of
the whole community. Content-based recommendation and collaborative
filtering are two mainstream collective view prediction techniques. They
generate predictions by analyzing the text features of the target object
or the similarity of users' past behaviors. Still, these techniques are
vulnerable to the artificially-injected noise data, because they are not
able to judge the reliability and credibility of the information
sources. Trust-based Collective View Prediction describes new
approaches for tackling this problem by utilizing users' trust
relationships from the perspectives of fundamental theory, trust-based
collective view prediction algorithms and real case studies.
The book consists of two main parts - a theoretical foundation and an
algorithmic study. The first part will review several basic concepts and
methods related to collective view prediction, such as state-of-the-art
recommender systems, sentimental analysis, collective view, trust
management, the Relationship of Collective View and Trustworthy, and
trust in collective view prediction. In the second part, the authors
present their models and algorithms based on a quantitative analysis of
more than 300 thousand users' data from popular product-reviewing
websites. They also introduce two new trust-based prediction algorithms,
one collaborative algorithm based on the second-order Markov random walk
model, and one Bayesian fitting model for combining multiple predictors.
The discussed concepts, developed algorithms, empirical results,
evaluation methodologies and the robust analysis framework described in
Trust-based Collective View Prediction will not only provide valuable
insights and findings to related research communities and peers, but
also showcase the great potential to encourage industries and business
partners to integrate these techniques into new applications.
