Quasi-Reflection Learning Arithmetic Firefly Search Optimization with Deep Learning-based Cyberbullying Detection on Social Networking

Abstract

Social networks are a major medium for communicating, collaborating, and sharing knowledge, data, and ideas. However, due to anonymity preservation, incidents of cyberbullying and hate speech emerge. Cyberbullying is very common on social media, and people end up with depression and do not take action against it. Automatic identification of these situations on many social networking sites requires intelligent systems. Deep learning (DL) methods are preferred for their potential in text classification, with accurate results on various academic benchmark issues. This study develops a Quasi-reflection Learning Arithmetic Firefly Search Optimization with Deep Learning Cyberbullying Detection (QLAFSO-DLCBD) technique to detect accurately cyberbullying on social media. The proposed QLAFSO-DLCBD method undergoes an initial preprocessing stage to convert the raw data into a meaningful format. The Keras embedding layer is used for word embedding purposes. The QLAFSO-DLCBD technique applies the Attention-based Bidirectional Long Short-Term Memory (ABiLSTM) method to detect cyberbullying. The QLAFSO algorithm was employed to select optimal hyperparameters for the ABiLSTM method, enhancing detection performance. Extensive experimental and comparative results suggest a higher efficacy of the proposed QLAFSO-DLCBD method compared to other recent methods.