A Novel Multimodal Hybrid Classifier Based Cyberbullying Detection for Social Media Platform

Platforms for social networking such as Facebook and Twitter, as well as others, provide a great number of benefits, but they also come with a great number of drawbacks. Cyberbullying is one of the problems that may occur on these social sites. The effect that cyberbullying has on the lives of its victims is incalculable. It is quite subjective, and the approach that each individual would take to this differs greatly. For those who are bullied, the message may seem natural, but for others, it may be intimidating. The ambiguity that may be found in cyberbullying texts makes it very difficult to locate the bullying material. The use of textual postings has been the subject of a significant amount of study, which has been recorded. In this work Extreme Learning Machine (ELM) based cyberbullying detection is proposed. Convolutional Neural Network (CNN) is used to classify emoji posted in the tweets. The text, hashtag, emoji datasets are preprocessed, and features are extracted. ELM classifier is used to detect the cyberbullying. This work achieves accuracy of 99.41, precision of 92.76, recall of 90.17, and f1-score of 91.64.