Deep Learning and Ensemble Machine Learning-Based Approach to Detect Cyberbullying Using Twitter Data

The development of the modern web as well as its widespread use on communication sites like Facebook, YouTube, and Twitter have made life significantly easier when handling some very private issues. The goal of the paper is to reduce online abuse and establish a criticism-free virtual community. In this research, we have assessed a significant number of English tweets using several characteristics. We use natural language processing and machine learning (ML) approaches to convey clean information. Furthermore, the group has expanded the use of deep learning (DL) methodologies by applying tokenization with padding. Harassment that occurs over the internet can be swiftly identified through the use of natural language processing as well as mathematical visualization. A variety of classifiers have been utilized in our study. We increased the study by including several DL structures. It takes a lot of information to accurately identify acts of harassment. To quickly identify articles related to online abuse, we integrated a pair of data sets, resulting in $1,47,682$ tweets about various viewpoints. We can notice which combination structure (MLP+SGD+LR) conducts better after comprehending the ML and DL approaches. Its multilabel assessments of accuracy, precision, and recall are all 99.11%.