Neural Network-Based Cyber-Bullying and Cyber-Aggression Detection Using Twitter(X) Text

Abstract

Cuber-bullying and Cuber-aggression analysis and detection form a critical aspect in today's social networking research being driven by machine learning and deep learning techniques.Quite a number of several techniques have been proposed to measure and detect unpleasant or offensive content or behavior on different social platforms over the years; this includes chat analysis of speech and fuzzy logic, natural language processing (NLP) among others.In this research, sentiment analysis of tweeter text using hybrid neural network to detect cyber-bullying and cyber-aggression from users' tweets on twitter and google images was developed with RNN for the text analysis and CNN for image analysis.The proposed methodology exploits sentiment analysis features like polarity of text, specific feelings and emotions, intentions and images to determine bullying or aggression.The text data is generated from Twitter through Twitter API and images from google.Deep learning models (RNN and CNN) continue to demonstrate their potential in the task of predictive analytics to tackle data analysis and learning challenges as the size of data grows and the need for a quick and accurate result grows.The efficiency and performance of models were evaluated with RNN and CNN outperforming the other classification algorithms achieving accuracy of 0.951 and 0.911 also, F-Measure scores were 0.910 and 0.890 showing impressive performance in the analysis and detection in text and images.This research provides a substantial contribution to cyberbullying and cyberaggression analysis and detection mechanism by tapping into Twitter users' psychological features including personalities, sentiment and emotion.domain by harnessing the potentials deep learning.