Detecting E-Bullying in Social Media Platform with Stacked BiLSTM Approach

Nowadays, E-bullying is a widespread phenomenon faced by people in the digital realm and has been escalating rapidly. E-bullying poses a significant threat to individual’s well-being and mental health that can even lead to suicide attempts. Detecting and addressing such behavior on social media platforms is crucial in ensuring a safe and supportive online environment. The Stacked BiLSTM architecture is presented for identifying E-bullying. The methodology encompasses data collection from Kaggle, followed by data cleaning, and preprocessing techniques to ensure dataset quality. Subsequently, the deep learning model is trained using TensorFlow and Keras libraries, leveraging the bidirectional processing and hierarchical representation learning capabilities of Stacked BiLSTM networks. The model’s effective-ness is assessed on a balanced dataset of 4000 texts, yielding a learning accuracy of 88.7% and an assessment accuracy of 82.2%. Result analysis underscores the efficacy of SBiLSTM to identify e-bullying, indicating potential for the development of improved detection systems. Overall, the proposed approach contributes to mitigating online harassment and fostering safer digital environments.