Cyberbullying Detection in Social Networks Using the Big Five Model

• Main Authors: Yu-An Lin, 林于安
• Other Authors: Yung-Ho Leu
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/03706141475059191325

碩士 === 國立臺灣科技大學 === 資訊管理系 === 104 === Due to popularity of the Internet and social media, people are eager to share their thoughts and feelings through the social media. However, the social media is a double-edged sword; it not only brought joys to the public for information sharing, it also brought agony to the public through cyber bullying. This research aims at identifying people who are suffering from cyber bullying by screening their posts on Instagram. To this end, we have built thesauruses for different personality traits. Then, we identified a user's personality trait by matching the contents of his posts on Instagram and the thesauruses, and used the personality trait to predict the possibility for the user being cyber bullied. The whole process has been performed in two steps. In the first step, we built four thesauruses according to the Big Five model and the posts on Instagram. Then, we conducted a survey by issuing questionnaires to the users of Instagram to collect a training dataset. The training dataset contains the personality traits and the yes's or no's for the users being cyber bullied. We used the training dataset to verify the accuracy on predicting a user's personality trait and to build a prediction model for cyber bullying. The contribution of this thesis includes (1) to use the posts in Hashtags and the Big Five model to construct thesauruses for personality traits, (2) to use the posts of a user to predict the user's personality trait, and (3) to use the predicted personality trait and other related attributes to predict whether or not a user is being cyber bullied. The experimental results showed that the precision and the F-Measure of the prediction are 0.8 and 0.727, respectively.