Matches in ScholarlyData for { <https://w3id.org/scholarlydata/inproceedings/www2012/poster/54> ?p ?o. }
Showing items 1 to 15 of
15
with 100 items per page.
- 54 creator guan-wang.
- 54 creator philip-s-yu.
- 54 creator shuyang-lin.
- 54 creator sihong-xie.
- 54 type InProceedings.
- 54 label "Review Spam Detection via Time Series Pattern Discovery".
- 54 sameAs 54.
- 54 abstract "Online reviews play a crucial role in today’s electronic com-merce. Due to the pervasive spam reviews, customers canbe misled to buy low-quality products, while decent storescan be defamed by malicious reviews. We observe that, inreality, a great portion (> 90% in the data we study) of thereviewers write only one review (singleton review). Thesereviews are so enormous in number that they can almostdetermine a store’s rating and impression. However, exist-ing methods ignore these reviewers. To address this prob-lem, we observe that the normal reviewers’ arrival patternis stable and uncorrelated to their rating pattern tempo-rally. In contrast, spam attacks are usually bursty and ei-ther positively or negatively correlated to the rating. Thus,we propose to detect such attacks via unusually correlatedtemporal patterns. We identify and construct multidimen-sional time series based on aggregate statistics, in order todepict and mine such correlation. Experimental results showthat the proposed method is effective in detecting singletonreview attacks. We discover that singleton review is a signif-icant source of spam reviews and largely affects the ratingsof online stores.".
- 54 hasAuthorList authorList.
- 54 isPartOf proceedings.
- 54 isPartOf proceedings.
- 54 keyword "Review spam".
- 54 keyword "Temporal pattern mining".
- 54 keyword "Time series".
- 54 title "Review Spam Detection via Time Series Pattern Discovery".