Cyberbullying (CB) continues to grow across social media platforms, creating a need for advanced and reliable detection systems to help protect online users. This study introduces a hybrid deep learning framework named DEA-RNN for detecting cyberbullying on the Twitter platform. The proposed model integrates an Elman Recurrent Neural Network (Elman RNN) with the Dolphin Echolocation Algorithm (DEA), which is employed to optimize the network’s parameters and minimize training time. To evaluate the effectiveness of DEA-RNN, experiments were conducted on a dataset of 10,000 tweets and its performance was benchmarked against several state-of-the-art classifiers, including Bi-directional Long Short-Term Memory (Bi-LSTM), standard RNN, Support Vector Machine (SVM), Multinomial Naive Bayes (MNB), and Random Forests (RF). The results demonstrate that DEA-RNN consistently outperforms all comparison models across multiple evaluation scenarios. Notably, in Scenario 3, DEA-RNN achieved superior performance with an average accuracy of 90.45%, precision of 89.52%, recall of 88.98%, F1-score of 89.25%, and specificity of 90.94%, highlighting its robustness and effectiveness for cyberbullying detection in social media environments.
