Detecting Predatory Behaviour in Online Game Chats

This paper describes a machine learning approach to detect
sexually predatory behaviour in the massively multiplayer online game for children, MovieStarPlanet. The goal of this work is to take a chat log as an input and outputs its label as either the predatory category or the non-predatory category. From the raw in-game chat logs provided by MovieStarPlanet, we first prepared three sub datasets via extensive preprocessing. Then, two machine learning algorithms, naive Bayes and Decision Tree, were employed to model the predatory behaviour using different feature sets. Our evaluation has revealed that the proposed
approach achieved high accuracies in detecting predatory chats