Ensemble Learning With Tournament Selected Glowworm Swarm Optimization Algorithm for Cyberbullying Detection on Social Media

Online social network (OSN) plays a crucial role to facilitate social connections; but, this social networking media increases antisocial behaviors, like trolling, cyberbullying, and hate speech. Cyberbullying has often resulted in serious physical and mental distress, especially for children and women, and even sometimes forces them to commit suicide. Conventional techniques for detecting cyberbullying, such as relying on users to report the instance of bullying, are not always effective. Deep learning (DL) and Machine learning (ML) techniques are trained to automatically recognize and flag potential cyberbullying content, along with identifying behavior patterns that are indicative of cyberbullying. Therefore, this study concentrates on the design and development of ensemble deep learning with tournament-selected glowworm swarm optimization (EDL-TSGSO) algorithm for cyberbullying detection and classification on Twitter data. The goal of the study is to examine social media data through the use of natural language processing (NLP) and ensemble learning process. This EDL-TSGSO technique preprocesses the raw tweets and then employs the Glove word embedding technique. In addition, the presented EDL-TSGSO technique utilizes ensemble long short-term memory with Adaboost (ELSTM-AB) model for effective cyberbullying detection and classification. The ensemble ELSTM-AB classifier integrates the prediction of LSTM and Adaboost models to enhance the overall classification performance. To further develop the cyberbullying detection performance of the EDL-TSGSO algorithm, the TSGSO algorithm is applied as a hyperparameter optimizer. The experimental validation of the EDL-TSGSO algorithm on the Twitter dataset demonstrates its promising performance over other state of art approaches in terms of different measures.