Social network approach for cyberbullying detection using machine learning

Cyberbullying is a serious issue that affects both adults and teenagers on the internet. Using social media and the internet is frequently associated with sending, receiving, and publishing derogatory, false, or unpleasant content about other people. This shows that cyberbullying has had a substantial negative impact on mental health, especially among the younger population. If action is not taken to stop cyberbullying, self-esteem and problems with mental health will impact a whole generation of young adults. Considering this, it is necessary to use machine learning (ML) approaches combined with natural language processing (NLP) and techniques like TF-IDF (Term Frequency-Inverse Document Frequency) to detect cyberbullying effectively. This study utilised a dataset that includes data on hate speech tweets. The research began by analysing the nature of cyberbullying and the challenges in its detection, underscoring the significance of automated methods. The study used NLP and TF-IDF to pre-process and analyse the dataset, identifying patterns and characteristics typical of cyberbullying behaviours. Subsequently, various ML techniques were utilised to accurately train models that can detect instances of cyberbullying in social media content. Specifically, the study has three primary goals: to create and implement an effective method for detecting online abusive and bullying messages by integrating NLP with ML; to evaluate the accuracy of the proposed detection algorithms for cyberbullying, specifically using Support Vector Machine (SVM), Naive Bayes (NB), and Random Forest (RF); and to compare the performance of these algorithms to identify the one with the highest accuracy in detecting text-based bullying. To sum up, this study highlights the potential of leveraging ML, NLP, and TF-IDF to address the escalating issue of cyberbullying on social media. The study advances the development of sophisticated detection algorithms by utilising a comprehensive dataset to emphasise the need for a multimodal approach that combines technological solutions with awareness-raising and education to foster a safer and more inclusive online community.