Bot Detection In Social Networks Using Random Forests (Bachelor Project, Finished)

Recent research indicates a considerable ratio of social bots automatically creating and publishing content in social networks. Due to the potential of abusing bots to promote certain content and influencing social media users, several frameworks claiming to be able to distinguish a social bot from a regular user have been developed. Here, we replicate the result of a previous study by Varol et al. which used the random forest algorithm to classify Twitter users. First, we look at relevant literature regarding the detection of social bots. After identifying relevant features of social network users to train our model we then evaluate several data sets of Twitter and Instagram users using our own implementation of a framework to predict user classifications. Our results show that random forests can indeed reliably separate legitimate users from bots for sufficiently large data sets and confirm the results previously reported by Varol et al.