Paper Info
Title
Detecting Social Spammers in Colombia 2014 Presidential Election.
Abstract
The large amount of user-generated content has turned social media into an appealing source of information for understanding social behavior. Around elections time, Twitter data have been used to measure public opinion on issues such as predicting outcomes, voting intention or political alignment. However, the effect of proliferation of new forms of spam on social media in this type of measurements has not been completely recognized and tackled in research. In this paper, we focus on detecting malicious accounts on Twitter, which aim to spread spam in an electoral process (e.g., disseminate rumors, misinform, or artificially inflate support for a candidate). To achieve this, a dataset of 149K users referring to Colombia 2014 presidential election was collected, and 1.7 million tweets and 341K URLs were crawled from their timeline. To distinguish malicious accounts from non-spammer ones in the dataset, several machine learning techniques were implemented on a labeled collection of users, semi-automatically classified into spammer and non-spammer. Experimental results reveal that with ten tweets, the proposed strategy detects 93% of spammers and 92% of non-spammers. Results also highlight the importance of noise removal when measurements of public opinion are conducted using Twitter data, with approximately 22% of accounts in the dataset classified as spammers.
Year
DOI
Venue
2015
10.1007/978-3-319-27101-9_9
ADVANCES IN ARTIFICIAL INTELLIGENCE AND ITS APPLICATIONS, MICAI 2015, PT II
Keywords
Field
DocType
Social spammers,Spammer detection,Twitter,Politics,Presidential election,Colombia
Internet privacy,Social media,Voting,Computer science,Presidential election,Timeline,Dissemination,Public opinion,Politics,Spamming
Conference
Volume
ISSN
Citations 
9414
0302-9743
2
PageRank 
References 
Authors
0.36
12
2
Name
Order
Citations
PageRank
Jhon Adrián Cerón-Guzmán130.76
Elizabeth Leon2335.26