A Machine Learning Approach for Detection of Fake Profiles on Twitter

Abstract

The advent of social media is an engineering marvel and a boon to mankind. With every advancement in technology, negative elements of the society are encountered who try to find loopholes and leverage them in order to victimize innocent users. In this context a similar problem is faced. The presence of fake profiles not only diminishes the experience offered by the platform but also possesses the power to cause heavy damage to an individual’s reputation and privacy. It is of paramount importance to find methods to arrest the same. In this project multi-variate analysis on freely accessible data is carried out from the user’s profile without any infringement of privacy and supervised learning algorithms for classification of profiles are applied. The goal is to track the typical behavior of fake profiles and contrast them with the characteristics of original profiles. Finally a comprehensive comparison of accuracy for the supervised learning algorithms and their respective relation to number of profile features so as to determine the best fit algorithm for the given data is provided.