Hybrid Deep Learning with DEA-RNN for Identifying Cyberbullying in Twitter Data

Abstract

Cyberbullying (CB) is on the rise in today's online communities. With so many people of all ages using social media, it's crucial that these sites be protected from harassment. In order to identify CB on the Twitter platform, this article introduces a mixed deep learning model dubbed DEA-RNN. To fine-tune the Elman RNN's characteristics and shorten training time, the suggested DEA-RNN model blends Elman type RNNs with an improved Dolphin Echolocation Algorithm (DEA). Using a dataset of 10,000 tweets, we conducted extensive testing on DEA-RNN and compared its results to those of other state-of-the-art algorithms like RNNs, SVMs, Multinomial Naive Bayes, and Random Forests. (RF). The testing findings indicate that DEA-RNN performs better than the alternatives in every situation tested. In terms of identifying CB on Twitter, it did better than the other methods that were taken into account. With an average of 90.45% accuracy, 89.52% precision, 88.98% memory, 89.25% F1-score, and 90.94% sensitivity, DEA-RNN performed best in case 3.