Bot Detection in Social Networks Using Stacked Generalization Ensemble

In recent times, the reach and influence of social media have grown tremendously across the entire globe. The ease of access, simplicity, publicity and reach offered by giant social networking sites have come to hold immense value nowadays. However, this has led to the widespread use of fake accounts or programmed bots in order to inflate one’s social media popularity and further spread favourable content. Many recent studies have highlighted the impact of such bots in fields like advertising, commercial promotion and even elections. In this paper, we propose a method to detect bots on social networking sites and distinguish them from genuine user accounts by using a stacked learning approach whereby a convolutional neural network model is trained to feed forward to a machine learning model. This is achieved by using a supervised learning approach to build a layered classifier that makes predictions based on a user’s profile information, tweets and activity information from a dataset of Twitter users. Our paper also analyses the comparative performance of many machine learning models applied to this problem.