The prevalence of cyberbullying (CB) in online entertainment settings is rising. Because social media is so widely used by people of all ages, it is imperative that the platforms be protected from cyberbullying. This paper introduces DEA-RNN, a hybrid deep learning algorithm for Twitter CB detection. Elman-type recurrent neural networks (RNNs) and an enhanced Dolphin Echolocation Algorithm (DEA) are combined in the suggested DEA-RNN model to shorten training times and optimize the parameters of the Elman RNNs. We thoroughly tested DEA-RNN using a dataset of 10,000 tweets and contrasted its results with those of state-of-the-art algorithms including Random Forests (RF), Bi-LSTM, RNN, SVM, and Multinomial Naive Bayes (MNB). The studies’ outcomes show that DEA-RNN performed better in every circumstance. It fared better than previously thought-of methods in terms of identifying CB on the Twitter website. 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%, DEA-RNN performed better in scenario 3.