A Live Suspicious Comments Detection using TF-IDF and Logistic Regression

Abstract

The internet has changed the lives of everyone since its arrival.  It gave birth to social networking platforms and online forums where people can share their thoughts and it also contains vast amount of information and it is become an effective and convenient communication tool for people to convey their thoughts. Although internet paves the way for many benefits but it also has its own drawbacks. One of the drawbacks is the threat of abuse and harassment for sharing our thoughts. Many platforms fail to moderate user comments and toxic behaviors which restrict people from expressing themselves. In this concern, we specialize in creating a machine learning model using logistic regression algorithm to implement the model in a live chat to detect abusive and harmful comments in real time. This research detects different types of toxicity levels and classifies them into toxic, threat, severe toxic, obscene, insults and identity-based hatred and to list out the names of the abusers and toxic users.