Scholarlink

Cyberbullying poses a significant challenge in today’s digital society due to its widespread reach and persistent effects. This paper investigates the role of metadata in enhancing cyberbullying detection across three datasets from YouTube, Twitter, and Sina Weibo. Using Correlation-Based Feature Selection and machine learning classifiers (Naive Bayes, Random Forest, and Bagging), the impact of user and social media meta-features on detection performance is analysed. The results show that while user metadata had minimal influence on detection accuracy, social media metadata, particularly, Retweets, Favourites, and SenderFollowers, significantly improved detection performance, enhancing accuracy by up to 9.7%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Power of Social Media Meta Data: Enhancing Cyberbullying Detection

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhou Zhou,
  • Tasmina Islam

摘要

Cyberbullying poses a significant challenge in today’s digital society due to its widespread reach and persistent effects. This paper investigates the role of metadata in enhancing cyberbullying detection across three datasets from YouTube, Twitter, and Sina Weibo. Using Correlation-Based Feature Selection and machine learning classifiers (Naive Bayes, Random Forest, and Bagging), the impact of user and social media meta-features on detection performance is analysed. The results show that while user metadata had minimal influence on detection accuracy, social media metadata, particularly, Retweets, Favourites, and SenderFollowers, significantly improved detection performance, enhancing accuracy by up to 9.7%.