Key points are not available for this paper at this time.
Abstract The use of information and communication technologies to engage in bullying behavior is known as cyberbullying. Today's society is facing a major and growing challenge of cyberbullying as a result of the misuse and inappropriate usage of social media. A few significant researches have been done in Artificial Intelligence (AI) inspired cyber bulling detection domain because of not having related dataset. This paper focuses on AI based cyber bullying detection in the context of social networking sites of Facebook, Twitter, Instagram, TikTok and YouTube English language. This paper has two major contributions. Firstly, we developed a dataset that involves collecting unique comments, evaluating them with psychological references, and categorizing them using Word Embedding for streamlined classification. Secondly, we offer a novel, machine learning-based solution to efficient cyberbullying detection systems which leverage the concept of advanced natural language processing techniques, including text preprocessing, feature extraction, and sentiment analysis, are employed to capture the intricate nuances of online interactions. Additionally, computer vision enhances detection beyond textual content. The methodology integrates various machine learning models, such as Logistic Regression, Decision Tree Classifier, Random Forest Classifier, Multinomial NB, KNeighbors Classifier, SVM, SGD Classifier, and Support Vector Machines. Experimental results, including Bidirectional LSTMs, showcase high accuracy, precision, recall, and F1-score metrics, demonstrating robust performance in handling diverse forms of cyberbullying and harassment. The paper concludes with insights into ethical considerations and future directions, highlighting the support vector machine (SVM) as the most effective algorithm with a 90.06% accuracy rate. Recommending SVM for social media platforms, the research contributes to enhancing online safety, guiding proactive measures against cyberbullying, and fostering a safer, more inclusive digital environment.
Islam et al. (Tue,) studied this question.