Cyberbullying Detection in Twitter Using Deep Learning Model Techniques

Abstract

The aim of the research is to develop a combined deep machine learning model (DEA_RNN) that will detect cyber bullying on Twitter. The suggested approach merges Elman-type Recurrent Neural Networks (RNNs) with a refined Dolphin Echolocation Algorithm (DEA) to improve parameter optimization and reduce the training duration. The proposed approach can handle the energetic nature of brief writings, adapt with point models for compelling extraction of trending themes and is proficient in recognizing and classifying cyberbullying tweets. The four modules in the proposed system are Dataset Collection, Data Cleaning and Preprocessing, Algorithm Implementation, and Prediction. The data cleaning and preprocessing phase involve noise removal, out-of-vocabulary cleansing, and tweet transformations to improve feature extraction and classification accuracy. The model algorithm uses DEA and RNN to generate a kind of echo similar to the behavior of dolphins during the hunting process. This research attempts to address the issue of recognizing cyberbullying on Twitter, which could be a challenging errand due to the energetic nature of brief writings and the changing nature of cyberbullying. The proposed approach beats models such as Simple Recurrent Neural Network (RNN), in terms of execution.