Organized behavior classification of tweet sets using supervised learning methods
8th International Conference on Web Intelligence, Mining and Semantics, WIMS 2018, Novi-Sad, Sırbistan, 25 - 27 Haziran 2018, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1145/3227609.3227665
- Basıldığı Şehir: Novi-Sad
- Basıldığı Ülke: Sırbistan
- Anahtar Kelimeler: 2016 US presidential elections, Big data, Organized behavior detection, Political propaganda, Social media analysis, Supervised learning, Twitter
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- Boğaziçi Üniversitesi Adresli: Evet
Özet
There is an increasing incidence in negative propaganda and fake news, which has recently gained lots of attention during the 2016 elections in United States, France, and United Kingdom. Bots and hired users collaborate to make messages seen and persist so they may spread and gain support. Assuming that most Twitter users post without predetermined, malicious intent, there is a need for automated detection of organized behavior to protect users from manipulation. This work proposes an automated approach to classify tweets with organized behavior. Supervised learning methods are used to classify the tweets by using a training data set with 850 records based on the analysis of over 200 million tweets. Our model gave promising results for detection of organized behavior and this motivated us to proceed with the generation of two more classifiers such as [“political”, “non-political”] and [“pro-Trump”, “pro-Hillary”,“neither"]. In each cases, the random forest algorithm consistently results in high scores with an average accuracy and f-measure above 0.95.