Logistic Regression Based Cyber Harassment Identification

Abstract

Increased online use and allowing users to engage with groups such as digital networking have contributed to the growth of hacking. Online abuse is a new type of harassment that has lately become more prevalent as online communities have grown in popularity. It tends to send messages which included defamatory claims or vocally harassing someone while in the internet group. Only if modern civilization recognizes harassment as it truly is, countless of hidden sufferers may continue to suffer. There have been several studies on cyber bullying, but none of them have been able to offer a solid remedy. By creating a model that can recognize and block bullying-related incoming and outgoing communications, we address this issue in our project. By employing supervised classification techniques on an open source dataset that has been carefully annotated, we hope to provide lexical baselines for this job. We have employed a logistic regression classifier for training and identifying instances of bullying behaviors. The dataset we used is a twitter dataset collected from kaggle. Our model classifies a message whether it’s bullying or not.