Show simple item record

dc.contributorCaverlee, James
dc.creatorFayazi, Amir A
dc.date.accessioned2014-05-13T17:21:44Z
dc.date.accessioned2017-04-07T20:07:14Z
dc.date.available2014-05-13T17:21:44Z
dc.date.available2017-04-07T20:07:14Z
dc.date.created2013-12
dc.date.issued2013-11-25
dc.identifier.urihttp://hdl.handle.net/1969.1/151744
dc.description.abstractUser submitted reviews are used by potential buyers to evaluate products before their purchase. In this work we study cases of deceptive reviews on Amazon.com which rate the products favorably. These were paid for through a number of crowd- sourcing websites. The behavior of the review spammers as a group has distinguish- able characteristics which are used in our proposed method. We use a probabilistic model for spammer pairwise collaboration which is used to cluster reviewers. The introduced model is verified on a set of synthetic data and outperforms a baseline classifier which treats reviews on their own, without their social context. The performance of the proposed method for detecting clusters of spammers is also compared to an alternative approach. Finally we demonstrate some of the detected clusters of review spammers on the data set which was crawled from Amazon.
dc.language.isoen
dc.subjectReview Spam
dc.subjectFake Reviews
dc.subjectSemi-Supervised Learning
dc.titleDetecting Crowdsourced Spam Reviews in Social Media
dc.typeThesis


Files in this item

FilesSizeFormatView

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record